skills-staging: session-derived campaign-ops skill library (2026-07-30) - #101
Merged
Merged
Conversation
Retiring-architect delivery: five skills plus START-HERE, MANIFEST, and UNCERTAINTY, distilled from the contributor-PR campaign session (#85-#97) with every repo-facing claim re-verified on 2026-07-30. Delivered by explicit force-add per the task's delivery model (skills-staging/ is gitignored by 92314a6; .gitignore untouched; nothing else under skills-staging/ is included). Three fresh-context reviews (factual/doctrine/usability) ran before delivery; all BLOCKING and IMPORTANT findings applied. Transcript-disposition risk filed as issue #100. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GhLu6NCGzM9HxnqNb87MAx
F-e-u-e-r
added a commit
that referenced
this pull request
Jul 31, 2026
…edger-retro/ Pure git mv, no content changes: the library's 11 files move from the skills-staging/ root into a dated namespace directory, matching the layout convention of the other two staged deliveries (#101, #106) so the staging zone stays one-directory-per-delivery. Path verified against the scoped-ignore whitelist landing via #107. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GhLu6NCGzM9HxnqNb87MAx
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What this is
Entry point:
skills-staging/2026-07-30-campaign-ops/START-HERE.md(nothing always-loaded points into a gitignored directory — this PR
is the pointer).
A project skill library distilled from the 2026-07-29/30 maintainer
session — the contributor-PR campaigns that produced #90 (
a148180),#93 (
92077a7), #97 (79ca49c) and issue #92 — authored under theretiring-architect protocol: extract rules from the session transcript,
verify every repo-facing claim against the repository, ship only what
verification supports, quarantine the rest in an uncertainty register.
All changes live under
skills-staging/2026-07-30-campaign-ops/:START-HERE.md— router: canonical-source map, current-state triage,precedence vs the older local library.
contribution-gate-playbook/— the combined-PR pipeline end to end(campaign-set pinning, ascending merges, integration fixes, gate
cadence, CI, owner-authorized merge, post-merge duties).
gate-adjudication-and-folds/— judging lens verdicts: reproducebefore folding, rejections with in-repo counter-evidence, executed
verification for runtime-semantics findings, opposite-fix
adjudication, TRIPLE PROCEED vs bounded-loop close.
doctrine-change-conventions/— covenant marker+Provenance pairing,line/glyph norms (and what checks.py does NOT check), Provenance
house style, the no-bump-for-doctrine version policy.
repo-boundaries-and-sync/— published vs private zones, thepermanently dirty-adjacent working tree, pathspec-only staging,
no-owner-quotes rule, the three-copy sync model and its current
expected state.
failure-archaeology/— the session's dead ends and residue withdisposition tags, plus the deliberately-not-done list.
MANIFEST.md— per-skill evidence and verification status.UNCERTAINTY.md— 9 entries, each with a safe default.What was extracted, and what was rejected or downgraded
Extracted from
chat-history.md(local untracked transcript of sessionsession_01WMbbZQ8ZMMsNRGyV2nvJR8— the id already stamped in thecampaign commits' trailers). The transcript was treated as evidence,
not ground truth; every repo-facing claim was re-verified on 2026-07-30
against: PR #90/#93/#97 bodies and merge commits, issue #92, evaluation
comments (sampled on #85/#94/#96), commits
f529014/814d116/6fa6154/5fa241b/83a038d/1560b97/32c6929/92314a6/8f8413f/c15c29d,.github/checks.pyandworkflows/checks.yml,.gitignore/git check-ignore, the version sites,git tag -l,live-install diffs,
git branch -vv, andgh pr viewreviewDecisionfields.
Rejected or downgraded during verification (details in UNCERTAINTY.md):
→ historically-valid only; Combined integration: #77-#80 + #82 through a seven-round tri-lens gate #84–Combined #94-#96: pairwise independence, single-arm replication, non-interactive-shell env + append-key clobber #97 doctrine merges kept
0.1.16.historically-valid; the current lineup ships date-stamped volatile.
evals/is gitignored with zero tracked files (UNCERTAINTY 8).skill-vettingexception the session itself disclosed(UNCERTAINTY 4).
holds across all three campaigns, but the governing rule text is
owner-set and session-recorded (UNCERTAINTY 2).
Delivery-model note (the gitignore interaction)
skills-staging/is gitignored by owner decision (92314a6,"retiring-architect scratch, not shipped"). This delivery follows the
explicit retiring-architect task instruction — deliver through GitHub
with all changes under
./skills-staging/— so the new subdirectorywas force-added;
.gitignoreitself is untouched, and nothing elseunder
skills-staging/(the 2026-07-12 local library) is included.Once merged, tracked files behave normally regardless of the ignore
rule. If keeping staging local is still preferred, closing this PR
loses nothing from the repo.
Reviews run before delivery
Three fresh-context reviews (factual / doctrine / usability) ran over
the staged files on 2026-07-30. Totals: 2 BLOCKING (one root cause),
11 IMPORTANT, 17 MINOR. Every BLOCKING and IMPORTANT finding was
applied before this PR opened. Notable dispositions:
issue was referenced before it existed — resolved by creating issue
Working tree carries untracked private session transcripts — disposition pending (gitignore pattern or relocation) #100 first and backfilling every reference, and by re-wording the
START-HERE triage so its counts stay true after this delivery.
git branch --mergedwas CONTRADICTED by observation in this repo(true merge commits; 37/37 local branches list as merged) and
rewritten to the observed mechanics; an unreproducible "36
occurrences" figure was replaced by a re-derived count (107
→hits in
skills/*/SKILL.mdat the pre-Combined #85-#89: probe results + trigger repair, bare-probe environment/tier/leakage/reasoning clauses, adoption + citation + staging rules, blanket-go scope, grader token collision #90 main); cadence quotesre-attributed to their real sources (Combined #94-#96: pairwise independence, single-arm replication, non-interactive-shell env + append-key clobber #97's cadence line lives in
its constituents' evaluation comments, not its body); the
remaining transcript-only specifics (the r8 close pre-declaration,
the fold-introduced attribution) are now labeled in place.
cross-model-review §4 (owner-surfaced close; runs past the
2–3-round cap only under the owner's standing mandate); the merge
step now requires an ask whose words cover MERGING plus the
CI-pending rule, and frames the observed review-bypass as a record,
not a norm; the House covenant is quoted verbatim ("probe or
trap"); the lens lineup is framed as a dated operating record with
cross-model-review §1 canonical; the old-library overlap map was
completed, including a frontmatter NAME COLLISION on
failure-archaeology between the two staged libraries.
installs
skill-vetting(a temporary per-directory overlay isdocumented, README named canonical); a description collision
between the two folding-related skills was resolved; the expected
sync state is single-homed in repo-boundaries-and-sync.
quotes, privacy trimming) except one accepted as-is:
failure-archaeology carries no Done definition (reference file;
its Re-verify serves that role).
Adjudications recorded on reviewer prompting: the House covenant's
probe/marker regime governs the pack roots — this staged library
ships under the MANIFEST/UNCERTAINTY regime instead, and this
delivery is not precedent for markerless pack rules; the lens lineup
appears here as an operating record, not pack text.
Validations performed
python3 .github/checks.pygreen with these files staged (thehidden-directive sweep covers every tracked file, now including
these).
descriptions are single-line per house style).
(cross-model-review §3/§4 headings; operational-rigor §2/§5).
affect them; CI runs both on the PR).
Remaining uncertainty and issues
UNCERTAINTY.mdcarries 9 entries, each with a safe default; theonly one needing an owner decision plus a change outside
skills-staging/is the untracked-transcripts disposition, filedas Working tree carries untracked private session transcripts — disposition pending (gitignore pattern or relocation) #100: Working tree carries untracked private session transcripts — disposition pending (gitignore pattern or relocation) #100
🤖 Generated with Claude Code
https://claude.ai/code/session_01GhLu6NCGzM9HxnqNb87MAx