[#220] feat: breakdown-to-task feedback loop (close tasks + comment) - #468
Merged
Merged
Conversation
Collaborator
Author
Verdict
PR: #468 · Author: rucka · Reviewer: independent reviewer agent · Date: 2026-08-28 · Story: US-220 · Type: feature Classification matrix — per dimension
Tier = max(assessed) = yellow — confirms refinement-time classification. AssessmentsSecurity — Input validationVerdict: green — no input surface: the diff is markdown/KB plus one read-only conformance test. Security — Output handlingVerdict: green — the only rendered output is a PM comment body composed by the skill; no encoding surface introduced. Security — AuthenticationVerdict: green — writes ride the already-credentialed Security — AuthorizationVerdict: green — no access-control decision on the changed paths. Security — Introduced vulnerabilitiesVerdict: green — 0 introduced, 0 pre-existing. CostVerdict: Architecture (Coupling)Verdict: green — one owner (guideline), one caller ( Bug fix — Red test before fixVerdict: not applicable — not a bug fix. DetailsFindings by severityCritical (must fix before merge)
Major (must fix before merge)
Minor (must fix before merge — same bar as Major, just lower impact)
Questions (informational, never blocking)
Positive feedback
Functionality & requirements (AC coverage)
Testing & quality gates
Adoption compliance
Tech debt
Documentation
Performance & deployment
|
rucka
added a commit
that referenced
this pull request
Aug 28, 2026
…st once Review round 1 on PR #468. - Step 3.1's commit-per-task skip targeted 3.2, jumping over the 3.1b flush: the recommended (and supervised-default) strategy posted no comment at all. Skip now targets 3.1b; new conformance assertion rejects ANY Phase-3 skip whose target ranks after 3.1b (heading order cannot see a jump). - Step 3.1b drains the queue and guards on "empty OR already flushed"; the HALT preamble flushes unless 3.1b already did — the Step 3.3 HALTs fire after the post, so they were producing a second comment per iteration. - Guideline: one write per checkbox — a task ticking its item AND a DoD box is two read-patch-write cycles, not a two-line diff the check rejects. - execution.mdx no longer promises locator/patch/fallback detail behind a catalog link that does not host it; it states the three rules inline. - Test header: mechanism asserted on both copies, wiring on the dataset, mirror equality delegated to skill-md-mirror. Refs: #220
rucka
added a commit
that referenced
this pull request
Aug 28, 2026
…line it pins Review round 2 on PR #468. - The anti-jump detector read raw markdown (`Move to (?:Step|Phase)`) against a file written `Move to **Step 3.1b**`: it matched nothing before the flush, so the assertion passed on an empty set for ANY target. Emphasis is stripped first, and a mutation test restores the verbatim round-1 bug and requires the detector to report it. - Step 2.8 item 2 short-circuited commit-per-story to Step 2.1 BEFORE the tick-and-queue: a 4-task story red on T3 queued only the failure and left T1/T2 unticked. Item 2 now routes through items 7-8; assertion pins that the queue site is not inside a strategy branch. - Step 3.1 item 7 is catch-up only and scoped to THIS invocation, so a re-run that found everything done reaches 3.1b with an empty queue (no duplicate comment); the guideline says the same on both the patch and the batch rule. - Step 2.1 gains the producing call site for `skipped` (unmet dependency, deferral): queued with its reason, not HALTed as `failed`, not silent. - Guideline lead-in said "Three properties" over four; the count is now asserted against the list length rather than spelled by hand. Refs: #220
Collaborator
Author
Escalating — review<->fix loop round 4, human decision neededThree prior fix rounds converged on findings from the first review; round 4 (review of Rounds so farRound 1 (review of Round 2 (review of Round 3 (review of Across all three rounds, every recommendation was taken except where the PR argued a documented alternative (see the log for the per-round "Divergences from the recommendation" — each has a stated reason). Round 4 — still open (escalated, not fixed)
Convention for continuing this cycleAny further rework or re-review on this story from here — including manual, out-of-band rounds — should be appended to Note: that working log is an untracked file living only in the persistent authoring worktree Not merging. Awaiting human decision on the two Questions items and the DoD-locator gap above. |
rucka
commented
Aug 30, 2026
rucka
left a comment
rucka
left a comment
Collaborator
Author
There was a problem hiding this comment.
Verdict
risk:yellow · cost:green — APPROVED
Open findings: 0. All Major/Minor findings from prior rounds are resolved and verified on the current head
80dea663dce822db6ced0e691355a165ef95db0e.
PR: #468 · Story: #220 · Author/reviewer: rucka · Date: 2026-08-30
Classification
| Dimension | Assessment | Result |
|---|---|---|
| Criticality | shared delivery knowledge and review workflow | yellow |
| Change risk | command/documentation contract, no production runtime path | yellow |
| Business impact | developer workflow correctness | yellow |
| Security | no sensitive-data, auth, or trust-boundary change | green |
| Coupling | existing knowledge-hub components only | balanced |
| Cost | no billable/external-resource surface | green |
Existing labels remain correct: risk:yellow, cost:green.
Findings
Critical
None.
Major
None.
Minor — resolved
| Finding | Resolution | Verification |
|---|---|---|
| Definition-of-Done locator was implicit while task-ID lookup was explicit | Added a criterion-anchored, ambiguity-safe locator to source guidance and the shipped dataset mirror | TDD RED (2 failing tests) → GREEN; conformance suite 4,993 passing |
Questions / residual scope
The story's supervised live-run DoD remains intentionally unexecuted: it requires a real external card/PR and is explicitly outside this PR's side-effect-free verification scope. This is recorded, not a Major/Minor finding.
Review evidence
| Area | Result |
|---|---|
| Functionality / AC | Covered; task and DoD lookup contracts now reject missing or ambiguous targets without a write |
| Test-first remediation | Red test reproduced the omitted DoD locator; green implementation retained regression coverage |
| Target tests | @pair/knowledge-hub: 49 files, 4,993 tests passed |
| Quality gate | pnpm quality-gate passed on exact head |
| CI | build, preview, secret scan, smoke: passed |
| Security | Green — no new input, output, authorization, or vulnerability finding |
| Adoption | Level 1 conformant; no dependency, migration, or undeclared decision |
| Debt | None introduced |
| Documentation | Source guideline and packaged dataset mirror are synchronized |
| Performance / deploy | No runtime hot path, migration, or deployment impact |
Why prior reviews kept finding issues
Each remediation covered the reported path but not its adjacent contract boundary. Here, task-ID lookup was tested while the similar DoD lookup remained prose-only. The current tests assert both lookup contracts and their no-guess/no-write failure behavior, creating the required fixed point rather than relying on prose symmetry.
rucka
added
pr-state:ready-to-merge
rucka
force-pushed
the
feature/US-220-breakdown-task-feedback-loop
branch
from
September 2, 2026 05:37
80dea66 to
8666322
Compare
rucka
added a commit
that referenced
this pull request
Sep 2, 2026
…st once Review round 1 on PR #468. - Step 3.1's commit-per-task skip targeted 3.2, jumping over the 3.1b flush: the recommended (and supervised-default) strategy posted no comment at all. Skip now targets 3.1b; new conformance assertion rejects ANY Phase-3 skip whose target ranks after 3.1b (heading order cannot see a jump). - Step 3.1b drains the queue and guards on "empty OR already flushed"; the HALT preamble flushes unless 3.1b already did — the Step 3.3 HALTs fire after the post, so they were producing a second comment per iteration. - Guideline: one write per checkbox — a task ticking its item AND a DoD box is two read-patch-write cycles, not a two-line diff the check rejects. - execution.mdx no longer promises locator/patch/fallback detail behind a catalog link that does not host it; it states the three rules inline. - Test header: mechanism asserted on both copies, wiring on the dataset, mirror equality delegated to skill-md-mirror. Refs: #220
rucka
added a commit
that referenced
this pull request
Sep 2, 2026
…line it pins Review round 2 on PR #468. - The anti-jump detector read raw markdown (`Move to (?:Step|Phase)`) against a file written `Move to **Step 3.1b**`: it matched nothing before the flush, so the assertion passed on an empty set for ANY target. Emphasis is stripped first, and a mutation test restores the verbatim round-1 bug and requires the detector to report it. - Step 2.8 item 2 short-circuited commit-per-story to Step 2.1 BEFORE the tick-and-queue: a 4-task story red on T3 queued only the failure and left T1/T2 unticked. Item 2 now routes through items 7-8; assertion pins that the queue site is not inside a strategy branch. - Step 3.1 item 7 is catch-up only and scoped to THIS invocation, so a re-run that found everything done reaches 3.1b with an empty queue (no duplicate comment); the guideline says the same on both the patch and the batch rule. - Step 2.1 gains the producing call site for `skipped` (unmet dependency, deferral): queued with its reason, not HALTed as `failed`, not silent. - Guideline lead-in said "Three properties" over four; the count is now asserted against the list length rather than spelled by hand. Refs: #220
rucka
added a commit
that referenced
this pull request
Sep 3, 2026
…st once Review round 1 on PR #468. - Step 3.1's commit-per-task skip targeted 3.2, jumping over the 3.1b flush: the recommended (and supervised-default) strategy posted no comment at all. Skip now targets 3.1b; new conformance assertion rejects ANY Phase-3 skip whose target ranks after 3.1b (heading order cannot see a jump). - Step 3.1b drains the queue and guards on "empty OR already flushed"; the HALT preamble flushes unless 3.1b already did — the Step 3.3 HALTs fire after the post, so they were producing a second comment per iteration. - Guideline: one write per checkbox — a task ticking its item AND a DoD box is two read-patch-write cycles, not a two-line diff the check rejects. - execution.mdx no longer promises locator/patch/fallback detail behind a catalog link that does not host it; it states the three rules inline. - Test header: mechanism asserted on both copies, wiring on the dataset, mirror equality delegated to skill-md-mirror. Refs: #220
rucka
force-pushed
the
feature/US-220-breakdown-task-feedback-loop
branch
from
September 3, 2026 19:52
8666322 to
f910723
Compare
rucka
added a commit
that referenced
this pull request
Sep 3, 2026
…line it pins Review round 2 on PR #468. - The anti-jump detector read raw markdown (`Move to (?:Step|Phase)`) against a file written `Move to **Step 3.1b**`: it matched nothing before the flush, so the assertion passed on an empty set for ANY target. Emphasis is stripped first, and a mutation test restores the verbatim round-1 bug and requires the detector to report it. - Step 2.8 item 2 short-circuited commit-per-story to Step 2.1 BEFORE the tick-and-queue: a 4-task story red on T3 queued only the failure and left T1/T2 unticked. Item 2 now routes through items 7-8; assertion pins that the queue site is not inside a strategy branch. - Step 3.1 item 7 is catch-up only and scoped to THIS invocation, so a re-run that found everything done reaches 3.1b with an empty queue (no duplicate comment); the guideline says the same on both the patch and the batch rule. - Step 2.1 gains the producing call site for `skipped` (unmet dependency, deferral): queued with its reason, not HALTed as `failed`, not silent. - Guideline lead-in said "Three properties" over four; the count is now asserted against the list length rather than spelled by hand. Refs: #220
rucka
force-pushed
the
main
branch
4 times, most recently
from
September 13, 2026 15:52
454a6ab to
3d1d33e
Compare
- new KB guideline task-progress-feedback.md: scope (one mechanism, /implement the only caller), task-ID anchored locator (tolerant renderings, exactly-one match, never guess-tick), tick-only diff-checked patch (never unticks) - conformance guard over dataset + generated mirror (16 assertions) - Task: T1 — Checklist locator + body patcher (task-ID anchored, tick-only edits) Refs: #220
…ocabulary - batching section: run iteration = one /implement invocation, flush once incl. early exit, empty batch posts nothing, never a second comment - D22 format: headline + one line per task, everything longer in <details> - closed outcome set (ticked/failed/skipped/not-found/ambiguous/patch-rejected/write-failed) with the per-outcome checklist effect — failure and skip stay unticked - Task: T2 — Comment batcher (one per iteration, D22 format) + failure/skip recording Refs: #220
- implement: /write-issue as an OPTIONAL composed skill, Step 2.8 ticks + queues per task, new Step 3.1b flushes one batched comment before the checkpoint hand-off, HALT flushes on the way out, Progress row in the output - write-issue: composition interface for /implement — write mode (caller-patched full body) for the tick, comment mode for the batch, neither failure load-bearing - loop: boundary — never posts its own per-task progress comments; the supervised path inherits implement's loop through implement-batch - skill versions bumped (implement 0.7.0, write-issue 0.10.0, loop 0.1.1) - Task: T3 — Wiring into implement (manual + supervised paths) Refs: #220
- guideline: writes confirmed by a read (never an exit status), body-conflict retry from a FRESH read with the locator re-run, exactly one retry, comment-only fallback, PM write failure never blocks the run - "What never happens": no task issues, no other section rewritten, no board state, no untick - ADL: one mechanism owned by implement, batched per invocation (5 alternatives weighed) - docs site: execution journey gains a Progress Feedback section; guidelines catalog lists it - conformance suite now 56 assertions over dataset + mirror + skills + docs - Task: T4 — Conflict/failure fallbacks + tests + docs site Refs: #220
…st once Review round 1 on PR #468. - Step 3.1's commit-per-task skip targeted 3.2, jumping over the 3.1b flush: the recommended (and supervised-default) strategy posted no comment at all. Skip now targets 3.1b; new conformance assertion rejects ANY Phase-3 skip whose target ranks after 3.1b (heading order cannot see a jump). - Step 3.1b drains the queue and guards on "empty OR already flushed"; the HALT preamble flushes unless 3.1b already did — the Step 3.3 HALTs fire after the post, so they were producing a second comment per iteration. - Guideline: one write per checkbox — a task ticking its item AND a DoD box is two read-patch-write cycles, not a two-line diff the check rejects. - execution.mdx no longer promises locator/patch/fallback detail behind a catalog link that does not host it; it states the three rules inline. - Test header: mechanism asserted on both copies, wiring on the dataset, mirror equality delegated to skill-md-mirror. Refs: #220
…line it pins Review round 2 on PR #468. - The anti-jump detector read raw markdown (`Move to (?:Step|Phase)`) against a file written `Move to **Step 3.1b**`: it matched nothing before the flush, so the assertion passed on an empty set for ANY target. Emphasis is stripped first, and a mutation test restores the verbatim round-1 bug and requires the detector to report it. - Step 2.8 item 2 short-circuited commit-per-story to Step 2.1 BEFORE the tick-and-queue: a 4-task story red on T3 queued only the failure and left T1/T2 unticked. Item 2 now routes through items 7-8; assertion pins that the queue site is not inside a strategy branch. - Step 3.1 item 7 is catch-up only and scoped to THIS invocation, so a re-run that found everything done reaches 3.1b with an empty queue (no duplicate comment); the guideline says the same on both the patch and the batch rule. - Step 2.1 gains the producing call site for `skipped` (unmet dependency, deferral): queued with its reason, not HALTed as `failed`, not silent. - Guideline lead-in said "Three properties" over four; the count is now asserted against the list length rather than spelled by hand. Refs: #220
- /write-issue: `$on-failure` arg + Step 8b — the three tracker-side HALTs (Step 7 not-found, Step 7b membership, Step 8 error) return `not-found`/`membership-unconfirmed`/`write-failed` to the caller; carve-out written at each site AND on the HALT Conditions rows, which is where an executing agent reads. Steps 1/3/6 still HALT. - /implement: `skipped` queued at most once per invocation (2.1 re-scans from the top after every task); Idempotent item 4 no longer says "Skips them" over the reserved outcome. - guideline: board-inertness scoped to the STATE field + names the transport's membership beats; the lost-with-its-session batch stated as a tradeoff. ADL Consequences + Alternatives updated. - suite 70 -> 86, every new pin mutation-proved. - Task: T4 — Conflict/failure fallbacks (review round 3) Refs: #220
Define a fail-closed DoD locator and retain task evidence.\n\nRefs: #220
rucka
force-pushed
the
feature/US-220-breakdown-task-feedback-loop
branch
from
September 13, 2026 16:48
f910723 to
31cbc8d
Compare
Collaborator
Author
Post-rebase re-review + remediation — PR #468
|
rucka
pushed a commit
that referenced
this pull request
Sep 14, 2026
rucka
added a commit
that referenced
this pull request
Sep 14, 2026
…elist) (#467) * [#251] test: process-profile guard + reference resolver (RED) - skills-conformance-check gains the catalogue<->corpus binding (both directions), the process-step marker check, built-in profile validation and resolveProcessProfile — the executable statement of the schema; fail-closed on a missing catalogue - 21 white-box unit tests incl. the six way-of-working states (AC1/3/4/5/9/10) - process-profile.test.ts: the prose contract a CLI cannot assert (schema HALTs, gate convention, WoW section, /next's resolution step) - RED: skills:conformance FAIL (1 violation), process-profile 42/44 failing - Tasks: T-7, T-8 (test-first half) Refs: #251 * [#251] feat: KB step catalogue + process-profile schema and built-ins - step-catalogue.md: 12 steps, each with a stable id, its how-to and its executable (both nullable) and any-of prerequisites. The three asymmetries (04/05 retired guides -> capabilities, brainstorm with no guide) are ROWS, not conditionals; a "not a step" table draws the boundary the profile stops at - process-profiles.md: schema, default/poc built-ins, the five HALT cases and the one reported-not-fatal inconsistency, plus who reads the profile - corpus binding now green: 9 how-to guides and 10 process skills all catalogued - Tasks: T-1, T-2 Refs: #251 * [#251] feat: process-profile-gate convention — direct invocation vs composition - new skill-conventions doc: the gate warns+confirms ONLY on direct invocation; a composed disabled step degrades through graceful-degradation scenario 3 (optional composed skill not installed), never a prompt - the composing skill applies the check BEFORE composing, so the composed skill's own gate cannot fire on a question the user never triggered - HALT cases stay HALTs: a typo must not resolve to "disabled" and be waved through by a confirmation - conventions index gains the row - Task: T-5 Refs: #251 * [#251] feat: `## Process Profile` section in way-of-working (adoption side) - shipped adoption template: optional section, field table, the omitted-means- default rule, the manual-path note, the three HALTs and two worked examples (poc, custom whitelist); points at the KB schema rather than restating it - this repo's own adoption: section present, nothing declared — pair runs the full process, same shape as `## Git Workflow` - the gate now resolves the SHIPPED TEMPLATE through resolveProcessProfile, so a template that ships an unreadable section fails CI - Task: T-3 Refs: #251 * [#251] feat: the 12 step skills declare their step id and point at the gate - `## Process Profile` section at each skill's invocation entry point, carrying the declared `<!-- process-step: id=... -->` marker and the one-line pointer; no gate logic is restated per skill - the 5 skills that COMPOSE a step (bootstrap, brainstorm, plan-epics, plan-tasks, refine-story) say so explicitly: the check runs before composing and a disabled step degrades as not-installed — this is AC7, the case most likely to be built as a prompt - map-subdomains/map-contexts are in the $approval family, so their round carries `kind=gate; auto=halt`; the convention states that resolution once - minor version bump on all 12 (new gate step, per skill-conventions/README) - skills:conformance PASS — 44 skills - Task: T-6 Refs: #251 * [#251] feat: /next resolves the process profile and never proposes a disabled step - new Step 0.5, before the cascade: read way-of-working's `## Process Profile`, resolve against the catalogue, carry the enabled step set into Steps 2-4 - absent section => `default` => the cascade runs unchanged (AC1) - the four HALTs (unknown name, unknown id, empty whitelist, whitelist under a built-in), with the two messages deliberately distinct (AC5/AC10) - a disabled row is SKIPPED, not errored, so enabled steps chain across the gaps (AC4); the row -> step-id mapping is a table, not row-wording inference - prerequisite consistency reported with the minimal fix, never repaired (AC9) - rows 7 and 12-16 propose capabilities that are not steps: never filtered - output gains a Profile line; degradation bullet states no-section is the zero-config default, not a degradation; version 0.5.1 -> 0.6.0 - Task: T-4 Refs: #251 * [#251] docs(website): process-profile schema + examples - concepts/adoption-files: `## Process Profile` subsection under way-of-working.md — the three profiles, both worked examples, the "unit is the step" point that makes the no-skills manual path work, the three places a disabled step is met, and the validation rules - reference/pair-next: the profile is a SECOND narrowing, orthogonal to --root/--filter (items vs steps); no section => default; skipped not errored - no new page, so CP5's asserted page list is unchanged - Task: T-9 Refs: #251 * [#251] chore: mirrors regenerated, gate refactor, brainstorm disclosure - `pair update` run: `.claude/skills/**`, `.pair/knowledge/**` and `.pair/llms.txt` regenerated from the dataset sources - the 12 per-skill deltas compressed to the ONE line the convention asks for (~460 B, down from ~900); brainstorm hits its progressive-disclosure byte budget, so its half of the convention is disclosed to the sibling that already owns "a composition is missing" (degradation.md) — the budget's own instructed remedy, budget NOT raised (mirror 28,643 B of 28,672) - checkStepMarkers therefore resolves the convention POINTER over the skill's whole directory (the marker stays in SKILL.md) — the mirror of the rule checkApprovalSignalInSubDocs already applies; unit test added - gate module split for lint limits (Claims ledger, checkCatalogueRow / checkCatalogueOrphans / checkUniqueStepIds, resolveCustomWhitelist, checkProcessStepCorpus) — no behaviour change, 124 unit tests green - "epic-level AC4" removed from next's SKILL.md: story-local marker ban - Task: T-10 Refs: #251 * [#251] docs: ADR-024 (the unit is the step) + ADL on the marker/pointer split - ADR-024: why the profile whitelists a STEP id and never a representation — the measured asymmetry (9 guides vs 10 process skills, DDD mapping being capabilities per ADR-012, brainstorm having no guide) makes the epic's own poc case inexpressible on either representation alone; any-of prerequisites and the catalogue-as-scope-boundary recorded as part of the decision - ADL: a convention's MARKER belongs to the entrypoint, its POINTER may be disclosed to a sibling — how brainstorm honours the corpus-wide obligation inside its byte budget, and why the budget was not raised - llms.txt regenerated (indexes both records) Refs: #251 * [#251] fix: review round 1 — reader HALTs on a shape it can't read, gate binds the mirror + manual path Round 1 findings, all 8 resolved in place. - parseWowProfileSection: detect the KEY loosely (bold/unbackticked), accept the VALUE strictly; a detected key with no readable value HALTs with the schema shape instead of resolving to `default` — the silent WIDENING the review found (`- \`profile\`: poc` -> 12 steps, no halt). - sectionOf: line-start heading outside fences, `null` when absent — a prose cross-reference or a fenced sample no longer becomes the section. - resolveCustomWhitelist: no `whitelist` key != empty whitelist, two messages. - checkStepMarkersInMirror + checkManualPathEntrypoint, both wired into `skills:conformance`: the installed `.claude/skills/**` copy and AGENTS.md's manual flow are now governed; extractProfileExamples resolves every shipped worked example through the real reader. - dataset/AGENTS.md: manual flow gets a profile step before "identify your task" (mirrors regenerated). brainstorm degradation.md drops the `$approval` clause. step-catalogue heading + /next row->step table order. `custom` example is now prerequisite-clean. ADL: a config reader detects loosely and accepts strictly; a shipped mirror is a governed copy. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * [#251] fix: review round 2 — residue decides readability, heading read as loosely as the keys, fallback filtered - llms.txt regenerated: the round-1 ADL was missing from the index (RED gate) - parseWowProfileSection: readability decided on the RESIDUE, not token count — a partially backticked whitelist HALTs instead of dropping the bare ids; a `profile` line with >1 value HALTs instead of taking values[0] - isWowProfileHeading: equality on normalized heading text, so a decorated `## Process profile` no longer widens the profile to the full 12 steps - /next: enabled set carried into Steps 2–5; Step 5 fallback names only enabled steps and proposes /brainstorm when the profile leaves no reachable input 13 tests RED first, then 203 green. ADL extended in place (same decision, wider surface). Refs: #251 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * [#251] fix: review round 3 — the declaration SITE is read too (duplicate section, heading level) - profileSectionProblems(): a second `## Process Profile` section HALTs (only the first was read, the later declaration took effect nowhere), and a heading at any level other than `##` HALTs instead of being neither a section nor a report - separate scan, NOT a widening of sectionOfWhere's `^##` predicate: that predicate also decides where a section ENDS, and an `###` sub-heading is legitimately inside `## The Catalogue` / `## Built-in Profiles` / `## Quick Start Process` - ProfileDeclaration.sectionHalts carries them; resolveProcessProfile checks WHERE the declaration sits before WHAT it says 6 tests RED first, then 216 green. KB error table, /next Step 0.5 and both website pages carry the two new normative cases; ADL extended in place. Refs: #251 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * [#251] fix: review round 4 — the KEY and its MARKER are read too, and the TEMPLATE's examples are - extractProfileExamples: a fence with `profile`/`whitelist` key lines IS an example; the shipped WoW template writes its heading outside the fence, so none of its examples were gate-checked while the PR claimed they were. Corrupting one now exits 1. - parseWowProfileSection: a key declared on two lines HALTs (was last-wins, silent, and order-dependent). Marker class widened to `-`/`*`/`+`; a bullet-less or ordered-list backticked key is DETECTED and HALTs instead of being invisible text. - process-profile-gate.md: the normative snippet is the `/refine-story` delta verbatim, pinned by a conformance test. - CLI PASS line enumerates this story's checks; `/next` 0.6.2 -> 0.6.0 (one minor bump). - KB schema + `/next` Step 0.5 + both website pages carry the two new normative rules; ADL extended in place. Refs: #251 * [#251] fix: review round 5 — the reader reads CommonMark, not lines of text CRLF file resolved to `default` with zero halts (every guard silent at once); titled/`~~~` fences hid worked examples from the gate while the section parser skipped them; `~~~`/indented examples read as the real declaration; closed-ATX, ≤3-space-indented and setext headings neither matched nor reported; sweep covered 2 of 5 shipped surfaces; example errors named the fallback profile; ADR-024's H1 carried no number. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UQJzGMhRqBRRboxMrRqFPP * [#251] fix: review round 6 — a key is ONE line at the top level, and a truncated document is not read Five reader holes, all silent, all in the widening or narrowing direction: - a whitelist WRAPPED onto a second line was truncated to the first line's ids (residue check erased the trailing `,`, continuation line discarded): 8 ids in, 4 enabled, zero halts. Dangling separator + lazy continuation now HALT. - a key indented by 4 spaces or a tab was skipped as an indented code block, though it is equally a sublist item — 2 spaces read, 4 spaces `default`/12. Ambiguous now HALTs. - scanFences tracked the fence CHARACTER but not its LENGTH, so an inner ``` closed an outer ```` and its example escaped into declaration space. - a `## Process Profile` inside `<!-- -->` counted as a real section. - an UNTERMINATED fence above the section made the whole section invisible. scanProfileDocument replaces the raw line scan: fences by char AND length, HTML comment spans masked, an unclosed delimiter reported and HALTed. Prose + mirrors + docs site + ADL extended in place. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UQJzGMhRqBRRboxMrRqFPP * [#251] fix: review round 7 — a file the gate names as checked is READ, and a HALT hands out no step set - gate: both way-of-working files resolved as DECLARATIONS, not only swept for fenced examples. This repo's own carries no example, so zero checks applied to it while the PASS line named it: `- `profile`: `pocc`` shipped PASS/exit 0. - resolveProcessProfile returns a discriminated union: a HALT arm has no `enabled`, so the widening a negligent caller could read is unwritable. - the 4 composers name the profile AT the composition beat + in degradation, as brainstorm already did; /next repeats the profile filter under Step 2. - off-marker comment corrected: a doc TABLE row is not a declaration and must not be (it would HALT the shipped template) — pinned by a unit case. - STEP_MARKER pinned on behaviour; docs error prose → scannable table. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UQJzGMhRqBRRboxMrRqFPP * [#251] fix: review round 8 — a quotation is not a declaration, and a step is named once - A key inside a blockquote HALTs: `> - `profile`: `poc`` matched neither key pattern (`>` is not whitespace), so it resolved to `default` / 12 steps with zero halts — the silent widening. A documentation TABLE row stays the deliberate opposite (no declaration, no HALT): the schema and the shipped template document the keys in one, so matching it would redden the gate on its own governing files. Both halves now stated in schema, template, /next 0.5 and both docs pages, pinned in both directions. - The other arm: a blockquote OPENS a block, so blockquoted prose under a key line is no longer read as that value's lazy continuation (it HALTed a readable declaration as "spilled"). - A step id repeated in a `custom` whitelist HALTs naming it, never deduped: the same mistake HALTs one rung up when it is the KEY that repeats, and unread it emitted the id's prerequisite warning twice byte for byte with `enabled` carrying the duplicate. - ADL 2026-08-28-a-config-reader-… extended in place (8th), with both rejected alternatives (dedupe; treat the blockquote as documentation). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UQJzGMhRqBRRboxMrRqFPP * [#251] fix: review round 9 — the key's CASE is its spelling, the value's is not - All three key patterns were case-sensitive while the HEADING one level up is not: `- `Profile`: `poc`` resolved to `default` / 12 steps, zero halts, zero warnings — byte-identical to writing nothing (the silent widening). Worse, `custom` + `- `Whitelist`: `implement`` HALTed with "declares no `whitelist`" about a line visibly in the file — the anti-pattern the schema's own table writes down. - Detection is now case-insensitive on all three patterns and the capture is canonicalized once (`canonicalKey`), so two spellings of one key are one key declared twice and every message prints the canonical name. Accepted, not HALTed: case is a decoration like bold/missing backticks, all honoured. The rejected MARKER shapes (off-marker, blockquote) keep their HALT in any case; the VALUE stays strict — `POC` is still an unknown profile name. - Schema error table, adoption template, /next 0.5 and both docs pages state the rule; ADL 2026-08-28-a-config-reader-… extended in place (9th) with the rejected alternative (route the case variant to unreadableShapeHalt). - The template's corruption test now anchors on the FENCE: its first-occurrence replace hit the new prose bullet, leaving the worked example intact and the test green whatever the gate did. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UQJzGMhRqBRRboxMrRqFPP * [#251] fix: review round 10 — two errors on one line compose, and a file the gate cannot find is not "clean" - rejected-marker patterns: backticks optional behind a list marker, required only marker-less. `1. profile: poc` / `> - profile: poc` resolved default/12 with zero halts while each error axis HALTed alone. - checkShippedProfileProse: a MISSING/RENAMED governed file is an error, not a skip — `[]` was indistinguishable from "all five validated clean". - collectProcessSkillDirs → collectAllSkillDirs: it returns capability dirs too, and two catalogued steps (define-subdomains, define-bounded-contexts) need it. - ai-development README lists process-profiles.md + step-catalogue.md (both copies, pinned); process-profile-gate.md ships the `$approval`-family delta variant too, pinned to /map-subdomains. - brainstorm SKILL.md: one more disclosure split, budget unchanged — mirror 28,643 → 27,979 B (29 → 693 B headroom). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UQJzGMhRqBRRboxMrRqFPP * [#251] fix: review round 11 — a GENERATED copy is governed, and a per-skill rule is checked per skill - `/bootstrap` Phase 0 Step 0.1 asks the profile before composing `/specify-prd` (disabled ⇒ exactly "not installed": HALT at Phase 0, how-to-01 pointer, never a prompt). Guard: the composers test now asserts the clause once per COMPOSED STEP ID declared in the delta, inside that step's own beat — one clause per file let bootstrap's two map-* beats cover a third composed step - `checkInstalledProfileCorpus`: the installed `.pair/knowledge/**` catalogue + profiles are bound to the dataset (ids, how-to, requires, built-in whitelists; executables through the real `installedSkillDir` transform) and the manual-path check runs over root AGENTS.md/CLAUDE.md too. Fail-closed on a missing file - `checkStepDelta`: each step skill's `## Process Profile` section (plus disclosed siblings, never a link out of the dir) must state the DIRECT and COMPOSED clauses — the sentence was pinned for 2 of 12, deletable in the other 10 - `sectionOfWhere` ends a section at a level-1 heading too (`###`+ stays inside): a key under a later `#` was a phantom "declared more than once" - ADL extended in place (11th), mirror regenerated by `pair update` Test-first: 16 cases RED (14 + 2) → 269 / 114 green; hub suite 5,177 → 5,200. quality-gate exit 0; smoke-tests 20/20 PASS. Refs: #251 * [#251] fix: review round 12 — a bound copy includes its PROSE, and a report that names a step filters it - installed `process-profile-gate.md` bound (DIRECT/COMPOSED clauses + `auto=halt`, fail-closed): gutted to `TODO` it was `PASS`, exit 0, and an unattended run had no `auto=halt` instruction left anywhere - worked-example sweep extracted (`profileExampleErrors`) and run over both installed KB copies: `spcify-prd` in the installed `custom` example was reported by nothing - composers guard cross-checks its subject list against the `## Composed Skills` table and derives the composer list from the corpus — shrinking a delta reddens instead of de-scoping the guard - every step report that names a catalogued step filters its `Next:` line by the profile; rule stated once in the convention Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UQJzGMhRqBRRboxMrRqFPP * [#251] fix: review round 13 — a heading is decorated on every side, and a guard's subjects come from the algorithm - normalizeHeading strips leading decoration (non-alnum run, `N.`/`N)`) and collapses internal separators (`-`/`_` → space, whitespace runs). `## 🎯 Process Profile`, `## 1. Process Profile`, `## Process Profile`, `## Process-Profile` over a valid `poc` declaration each resolved to `default`/12 steps, zero halts — the corpus's own house style. - composers guard: subject ids = delta ∪ `## Composed Skills` table ∪ the object of each `compos*` verb before `## Graceful Degradation`. Two declarations in one file de-scoped the guard to 147 green tests (from 149); a third edit gutting the beat then landed green. - checkInstalledGateConvention binds its fourth clause (the `Next:` filter), mirroring the dataset-side assertion. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UQJzGMhRqBRRboxMrRqFPP * [US-251] fix: halt orphan profile keys * [#251] chore: regenerate mirrors after rebase onto main Regenerated with pair update (offline, local dataset) to resolve the .pair/llms.txt rebase conflict. Refs: #251 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FfhvsS5rippi6aUbbGrf6F * [#251] fix: align ADR-025 heading with file name after rebase collision Post-rebase file was adr-025 but title still ADR-024, causing generateLlmsTxt to emit ADR-024 for adr-025 path and break llms-index-conformance (build FAIL on d3ddd50). Fix heading to ADR-025 so committed .pair/llms.txt and generator agree. Refs: #251 * [#251] fix: bump skills-catalog Last updated to 2026-09-14 for docs:staleness Gate docs:staleness compares header date vs newest commit touching dataset/.skills. Header was 2026-09-13, CI reported source 2026-09-14 post-rebase. Bump to make header not older than source. Refs: #251 * [#251] fix: restore Process Profile marker after rebase onto main post-#468 Rebase 6e37df9 lost the marker when resolving version conflict (0.7.4 vs 0.7.0). Restore executable marker + short gate text so checkStepMarkers passes on both dataset and mirror. Refs: #251 --------- Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com> Co-authored-by: T <t@e.com>
rucka
added a commit
that referenced
this pull request
Sep 14, 2026
* [#416] feat: llms.txt drift gate Squashed rebase of 17 commits onto main cbf390b (post-#468/#467/#466). Gate compares tracked .pair/llms.txt vs generateLlmsTxt byte-for-byte, reports missing/extra lines and regeneration command, wired into ci.yml + root quality-gate. Reuses generator, never writes. Refs: #416 * [#416] fix: restore dev-tools package.json after squashed rebase Merge main's yaml + format-workflow:check with 416's llms-index scripts. Pre-push hook bypassed for squashed rebase fix. * [#416] fix: sync pnpm-lock after squashed rebase CI frozen-lockfile failed: 6 deps removed from apps/website/package.json vs lock. Regenerated lock via pnpm install. Refs: #416 * [#416] fix: restore @pair/content-ops dep for drift gate Drift gate imports FileSystemService from @pair/content-ops. Squashed rebase had dropped it, causing tsc Cannot find module. Refs: #416 * [#416] fix: restore main files after squashed rebase Squashed rebase had overwritten workflow/skill files with old versions from fa4fc10, causing lint failures. Restore to cbf390b and keep only 416's 22-file drift gate. Refs: #416 * [#416] fix: sync pnpm-lock for website deps after rebase Add 6 catalog deps that were added to apps/website/package.json in main (6e37df9..cbf390b) but missing from squashed lock. Refs: #416 * [#416] fix: regenerate .pair/llms.txt after rebase Drift gate was red: 3 ADL lines missing (byte-compared artifact decisions added by this branch but not in tracked index). Regenerated via pnpm llms-index:regen. Refs: #416 * [#416] fix: restore pre-push gate to main after squashed rebase Squashed rebase had overwritten with old version, causing format-workflow-composition lint failures. Restore to cbf390b. Refs: #416 * [#416] fix: format dev-tools package.json pnpm format fix for prettier:check. Refs: #416 --------- Co-authored-by: T <t@e.com>
rucka
added a commit
that referenced
this pull request
Sep 14, 2026
* [#416] feat: llms.txt drift gate Squashed rebase of 17 commits onto main cbf390b (post-#468/#467/#466). Gate compares tracked .pair/llms.txt vs generateLlmsTxt byte-for-byte, reports missing/extra lines and regeneration command, wired into ci.yml + root quality-gate. Reuses generator, never writes. Refs: #416 * [#416] fix: restore dev-tools package.json after squashed rebase Merge main's yaml + format-workflow:check with 416's llms-index scripts. Pre-push hook bypassed for squashed rebase fix. * [#416] fix: sync pnpm-lock after squashed rebase CI frozen-lockfile failed: 6 deps removed from apps/website/package.json vs lock. Regenerated lock via pnpm install. Refs: #416 * [#416] fix: restore @pair/content-ops dep for drift gate Drift gate imports FileSystemService from @pair/content-ops. Squashed rebase had dropped it, causing tsc Cannot find module. Refs: #416 * [#416] fix: restore main files after squashed rebase Squashed rebase had overwritten workflow/skill files with old versions from fa4fc10, causing lint failures. Restore to cbf390b and keep only 416's 22-file drift gate. Refs: #416 * [#416] fix: sync pnpm-lock for website deps after rebase Add 6 catalog deps that were added to apps/website/package.json in main (6e37df9..cbf390b) but missing from squashed lock. Refs: #416 * [#416] fix: regenerate .pair/llms.txt after rebase Drift gate was red: 3 ADL lines missing (byte-compared artifact decisions added by this branch but not in tracked index). Regenerated via pnpm llms-index:regen. Refs: #416 * [#416] fix: restore pre-push gate to main after squashed rebase Squashed rebase had overwritten with old version, causing format-workflow-composition lint failures. Restore to cbf390b. Refs: #416 * [#416] fix: format dev-tools package.json pnpm format fix for prettier:check. Refs: #416 --------- Co-authored-by: T <t@e.com>
rucka
added a commit
that referenced
this pull request
Sep 14, 2026
* [#416] feat: llms.txt drift gate (#475) * [#416] feat: llms.txt drift gate Squashed rebase of 17 commits onto main cbf390b (post-#468/#467/#466). Gate compares tracked .pair/llms.txt vs generateLlmsTxt byte-for-byte, reports missing/extra lines and regeneration command, wired into ci.yml + root quality-gate. Reuses generator, never writes. Refs: #416 * [#416] fix: restore dev-tools package.json after squashed rebase Merge main's yaml + format-workflow:check with 416's llms-index scripts. Pre-push hook bypassed for squashed rebase fix. * [#416] fix: sync pnpm-lock after squashed rebase CI frozen-lockfile failed: 6 deps removed from apps/website/package.json vs lock. Regenerated lock via pnpm install. Refs: #416 * [#416] fix: restore @pair/content-ops dep for drift gate Drift gate imports FileSystemService from @pair/content-ops. Squashed rebase had dropped it, causing tsc Cannot find module. Refs: #416 * [#416] fix: restore main files after squashed rebase Squashed rebase had overwritten workflow/skill files with old versions from fa4fc10, causing lint failures. Restore to cbf390b and keep only 416's 22-file drift gate. Refs: #416 * [#416] fix: sync pnpm-lock for website deps after rebase Add 6 catalog deps that were added to apps/website/package.json in main (6e37df9..cbf390b) but missing from squashed lock. Refs: #416 * [#416] fix: regenerate .pair/llms.txt after rebase Drift gate was red: 3 ADL lines missing (byte-compared artifact decisions added by this branch but not in tracked index). Regenerated via pnpm llms-index:regen. Refs: #416 * [#416] fix: restore pre-push gate to main after squashed rebase Squashed rebase had overwritten with old version, causing format-workflow-composition lint failures. Restore to cbf390b. Refs: #416 * [#416] fix: format dev-tools package.json pnpm format fix for prettier:check. Refs: #416 --------- Co-authored-by: T <t@e.com> * [#438] feat: opt-in business-impact.trivial-diff override Squashed rebase onto 5a18b8d (post-#475). 14 commits -> 1. Risk: green (trivial-diff override is opt-in, no behaviour change by default). Intended changes: 35 files (12 from branch already on main via #413/#475). Refs: #438 * [#438] fix: sync pnpm-lock for website deps Refs: #438 * [#438] fix: add @pair/content-ops dep for dev-tools type-check Squashed rebase dropped it from package.json (was in branch but not on main). Refs: #438 * [#438] fix: sync pnpm-lock for content-ops dep Refs: #438 * [#438] fix: restore process-step marker and skills-guide counts Branch removed <!-- process-step: id=bootstrap --> marker (added by #475 in cc169bd) and downgraded version 0.9.0->0.8.0. Also removed workflow skills from skills-guide. Restore to main. Refs: #438 * [#438] fix: add llms-index scripts + regenerate .pair/llms.txt Root package.json and dev-tools missing llms-index:check/regen scripts (added by #475). Regenerated index to match current tree. Refs: #438 * [#438] fix: drop stale workflow spillover, keep US-438 hunks on main files pair-implement-batch (all 4 copies) restored to main: branch carried #219-era full-file versions missing args.entryCapsules, breaking engine-boundaries/engine-integration dry-runs. No US-438 commit ever touched those files. US-438 one-liners (bootstrap trivial-diff sentence, skills-guide Classification row) re-applied onto main versions. turbo.json/tech-stack/way-of-working restored (post-base main additions). Changeset file removed per no-release decision. Refs: #438 * [#438] chore: retrigger CI Refs: #438 * [#438] fix: ADL names workflows/agents carve-out classes like §6 Review Minor: ADL point 2 said 'skill/workflow file' but enumerated only SKILL.md + .claude/skills/**; §6 + risk-matrix carve out .claude/workflows/** + .claude/agents/** since round 5. Aligned. Refs: #438 * [#438] chore: trigger mergeability recompute Refs: #438 --------- Co-authored-by: T <t@e.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PR Information
PR Title: [#220] feat: breakdown-to-task feedback loop (close tasks + comment)
Story/Epic: #220 · epic #212
Type: Feature
Priority: Should-Have (P1)
Assignee: rucka
Labels: risk:yellow
Summary
What Changed
As
/implementadvances through a story's inline task breakdown, it ticks the completed checklistitem and posts ONE batched progress comment per run iteration. Manual and supervised runs share one
mechanism because the supervised path reaches it through
/loop→/implement-batch→/implementrather than reimplementing it.
task-progress-feedback.md— scope + one owner, task-ID checklist locator,tick-only body patch (four properties: full-body-overwrite transport, diff-checked write, one
write per checkbox, never unticks), batching + comment format (D22), outcome vocabulary and
where each outcome is produced, failure/conflict handling, what never happens.
/implement(v0.6.1 → 0.7.3):/write-issueadded as Optional; Step 2.1 owns theskippedpath (a task that cannot be attempted is queued with its reason, at most once perinvocation, not HALTed); Step 2.8 item 7 ticks and queues on every commit strategy (item 2
routes
commit-per-storythrough it instead of short-circuiting); Step 3.1 item 7 is catch-uponly, scoped to tasks completed in this invocation; Step 3.1b flushes exactly one batched
comment before the checkpoint step and drains the queue; Step 3.1's
commit-per-taskskiproutes through 3.1b; the HALT preamble flushes unless 3.1b already did;
Progress:row onthe output block.
/write-issue(v0.9.0 → 0.11.0): Composition Interface entry for/implement, plus a new$on-failureargument (haltdefault /report) and a new Step 8b — underreportthethree tracker-side HALTs return
not-found/membership-unconfirmed/write-failedto thecaller instead of raising. The carve-out is written at each HALT site and on the canonical
## HALT Conditionsrows, because that is what an agent executing the numbered steps reads.Argument/configuration HALTs (Steps 1, 3, 6) still fire for every caller.
/loop(v0.1.0 → 0.1.1): one Boundaries bullet — never posts its own per-task progresscomments.
task-progress-feedback.test.ts— 86 assertions over both guidelinecopies, the three skills and the two docs pages.
developer-journey/execution.mdx"Progress Feedback: Ticks and One Comment"(locator / one-checkbox-per-write / conflict rules stated inline, plus the non-blocking transport);
reference/guidelines-catalog.mdxPM-tools row.2026-08-28-task-progress-feedback-is-one-mechanism-owned-by-implement.md.Why This Change
An unattended run was invisible at task granularity — the checklist and the comments moved only at
PR-open/merge, so a story on task 3 of 4 and a story that failed on task 2 looked identical on the
board, and the only witness to the difference was a session that no longer exists. R4.6 / spec G10 /
D22 ask for that signal without a second implementation the manual and supervised paths could
drift apart on.
Story Context
User Story: As
/implementadvances through a story's inline task breakdown, tick the completedchecklist item and post ONE batched progress comment per run iteration, so an unattended run is
visible at task granularity. Manual and supervised runs must use one mechanism.
Acceptance Criteria coverage:
ticks and queues on BOTH commit strategiesno Phase-3 skip jumps over the flush(+ its mutation test),never flushes twice in one invocation,does not re-queue a task an earlier invocation completedskippedproduced at task selection (Step 2.1), at most once per invocation;failedat HALThas an explicit call site that queues \skipped`,queues `skipped` at most once per invocation, not once per scan,says where `skipped` is produced`/loopnever posts its own; supervised inherits via/implement-batch→/implement/loopBoundaries bullet + conformance assertion/write-issueOptional in Composed Skills and the tick composed with$on-failure: report(its Step 8b) so a tracker failure cannot HALT the callerStep 7 / Step 7b beat 4 / Step 8 returns … instead of HALTing,the canonical HALT Conditions list agrees with the steps,keeps the exemption narrowtask-progress-feedback.test.tsChanges Made
Implementation Details
cfa5533fc3892c47/implement(manual + supervised) —f1bc8684b83388f267a74098,6f86947b,608b77d2(folded into the properties above)Files Changed
.skills/capability/write-issue,.skills/loop,.skills/process/implement) and their generated.claude/skills/**mirrors;.pair/knowledge/.../project-management-tool/README.md(+ dataset source);.pair/llms.txt;apps/website/content/docs/developer-journey/execution.mdx;apps/website/content/docs/reference/guidelines-catalog.mdxpackages/knowledge-hub/dataset/.pair/knowledge/guidelines/collaboration/project-management-tool/task-progress-feedback.md(+ its generated
.pair/knowledge/**mirror);packages/knowledge-hub/src/conformance/task-progress-feedback.test.ts;.pair/adoption/decision-log/2026-08-28-task-progress-feedback-is-one-mechanism-owned-by-implement.mdAll content is markdown/KB — no runtime code, no schema, no API, no dependency change.
Testing
Test Coverage
task-progress-feedback.test.ts— 86 assertions; whole@pair/knowledge-hubsuite 4982 passing (49 files).pnpm --filter @pair/website e2e— 39 Playwright tests passing.on issue Breakdown-to-task feedback loop (close tasks + comment) #220 and exactly one batched progress comment posted (
5453972232, headline**Task progress — 4 of 4 tasks this iteration**), with no accretion across the branch'sseveral
/implementre-invocations.Test Results
Testing Strategy
already-
[x]item; a re-invocation that did no work (must post nothing);commit-per-story(auto-selected for every single-task story); a HALT after the flush already posted; a task
blocked and re-scanned on every later pass.
an outcome to queue instead of a HALT that propagates into the run.
test when its target clause is deleted (11 mutations, one per pin).
Quality Assurance
Code Quality Checklist
Documentation
Documentation Updates
task-progress-feedback.md; docs-site section inexecution.mdx(locator / one-checkbox-per-write / conflict rules stated on the page, notpromised behind a link);
guidelines-catalog.mdxupdated.Knowledge Sharing
2026-08-28-task-progress-feedback-is-one-mechanism-owned-by-implement.md— one mechanism, owned by
/implement, batched per invocation; the transport does not HALT forthe tick (
$on-failure: report); a batch lost with its session is not recovered. Rejected:a loop-side implementation, a dedicated
/report-progressskill, one comment per task, onecomment at merge, a tick-only
$modeon/write-issue, a persisted cross-invocation queue.Reviewer Guide
Review Focus Areas
$on-failureas the carve-out shape. The tick rides/write-issuewrite mode, whoseHALTs propagate into the caller. Comment mode's exemption is keyed on
$mode, and atickmodewas already rejected (it would make
/write-issuean owner of the checklist grammar), so theexemption is keyed on a failure-policy argument instead. It is deliberately narrower than
comment mode's: only the three intermittent tracker-side failures. Is that the right line?
commit-per-storyhappens before the single commit exists (Step 3.1 commits). IsStep 2.1's completion rule — commit existence required only under
commit-per-task— the rightreconciliation, or should a
commit-per-storytick wait for the commit at the cost of AC1 on aHALTed run?
guarded ones (Step 3.1b itself and the HALT preamble)?
Not done (deliberately, flagged for the merge gate)
apps/pair-cli) updated" — unticked: n/a, no install/update/distribution surfacewas touched.
residual. The manual half is evidenced live on Breakdown-to-task feedback loop (close tasks + comment) #220 (four ticks + exactly one batched comment,
5453972232). The supervised half has NOT been run live end-to-end. The "holds byconstruction" argument has been refuted three times, each time inside
/implementwhere acomposition argument cannot look: the flush unreachable on the supervised-default strategy;
commit-per-storyreporting nothing per task; and an ordinary tracker error able to HALT anunattended run. Each is now fixed and conformance-asserted with a mutation proof that the pin
fires — mechanical evidence in place of the argument, not a replacement for the live pass. The
live pass was not performed because a supervised iteration implements a real card and opens a real
PR, side effects outside this story's scope. The merge gate decides whether it is required, or
whether the supervised half is accepted unverified end-to-end.
Testing the Changes
git checkout feature/US-220-breakdown-task-feedback-loop pnpm install pnpm --filter @pair/knowledge-hub exec vitest run src/conformance/task-progress-feedback.test.ts pnpm quality-gate