CHANGELOG.md

Changelog

All notable changes to the debugging plugin are documented here. Format follows Keep a Changelog; this plugin uses semantic versioning.

[0.8.5] - 2026-10-04

Added

• debugging ships a claude plugin eval suite for /debugging:debug. The seven skill-creator cases in skills/debug/evals/evals.json are converted to case directories under evals/ (tag converted), and three new cases (tag phase1-red-run) check that Phase 1 runs a reproduction command and observes its failing output before the first edit. Graders are deterministic where the behavior allows (regex, tool_order, tool_used, file-content regex), with a judge rubric beside them where the expectation is a judgment; every grader has pass and fail samples. The Bash-granting cases need --scaffold and --allow-tools Bash,Edit,Write. No skill text changes.

[0.8.4] - 2026-10-04

Changed

• /debugging:debug names its successors (#6279). A new ## Next section routes a landed fix with its regression test to /review:quality-gate, a design-flaw root cause to /planning:plan, and a failure no loop could reproduce to /bugs:write --file. The redesign escalation now names /planning:plan when the planning plugin is installed, with a generic fallback otherwise.

[0.8.3] - 2026-10-04

Changed

• The shipped recommendation-basis contract (context/recommendation-basis.md) follows the convention's 2.0.0 grounding bar: a design pattern is grounded in the canonical source that defines it, not in how popular it is; recency never discounts a canonical pattern definition; and a pattern found in a template, sample, or popular repository is checked against the principle it claims to serve.

[0.8.2] - 2026-10-04

Fixed

• Shared view-runtime.js synced (#6173): pages built by /debugging:debug (post-mortem) no longer say they saved a file. The claude.ai artifact viewer blocks a download the page starts itself without an error the page can see, so the page reported a save that never happened. The "Save as file" button now appears only on a page opened from file:// or served by the session bridge on 127.0.0.1, and after a download attempt the status reads "If no file was saved, select the text below and copy it." "Copy reply" and the selectable payload text are unchanged.

[0.8.1] - 2026-10-03

Changed

• Shared view-builder.mjs, view-runtime.js synced (#5868); no change to this plugin's own behavior.

[0.8.0] - 2026-10-03

Added

• Interactive post-mortem view, built with the shared view builder (#5864). /debugging:debug offers a view of its Phase 6 post-mortem: the loop, cause, fix and seam, the hypotheses with their verdicts and evidence, a tick for each the reader would re-open, and a copy-out of the challenge. scripts/build-view.mjs fills a checked-in template with the session's JSON as escaped data through lib/view-builder.mjs and lib/view-runtime.js, which the plugin now carries as generated copies with lib/html-escape.mjs. No page carries model-written markup or script, so log and error text stays data, and no page sits beside the post-mortem, which stays the record. The publish destination comes from the medium key of the rendered-views cascade (file when unset); the procedure is in skills/debug/reference/rendered-view.md.

[0.7.14] - 2026-10-02

Fixed

• plugin.json no longer sets $schema. claude.ai's marketplace sync stripped it with a warning, and Claude Code ignores it at load time.

[0.7.13] - 2026-10-01

Changed

• References to the claude-config, claude-memory and claude-ops plugins now use their new names, harness-config, harness-memory and harness-ops.

[0.7.12] - 2026-09-30

Changed

• The debug Boundary bullet in debug no longer asserts that the bundled skill ships with Claude Code. It keeps the provenance class, what the skill does and how it is invoked, in the native-references template form.

[0.7.11] - 2026-09-29

Added

• debug carries a Boundary section for the bundled debug skill. The bundled skill debugs Claude Code itself and is reserved for the person to run, so the model offers /debug when the problem is Claude Code rather than the user's application.

[0.7.10] - 2026-09-29

Changed

• debug Phase 5 takes the regression test's expected value from the bug report, never from what the fixed code returns.

[0.7.9] - 2026-09-28

Changed

• Argument hints on debug stay inside the 100-character house style (#3542). Examples, defaults, and flag catalogs that exceeded the budget now live in the skill body.

[0.7.8] - 2026-09-28

Changed

• context/recommendation-basis.md names the full convention by its path in the marketplace repository instead of an org-specific URL.

[0.7.7] - 2026-09-28

Changed

• debug's Phase 6 architectural recommendation carries a Basis: and, when it touches shared code, is grounded in its consumers first, per the recommendation-basis convention.
• Ships context/recommendation-basis.md, a byte-identical copy of the discipline recommendation-basis contract, since an installed plugin cannot read the repository's docs/. An unsettled consequential recommendation is withheld and filed as an open question.

[0.7.6] - 2026-09-21

Changed

• American spellings throughout this plugin's prose, ahead of the en-us locale the shared typos config adopts. Wording only: no behavior, option, default, or identifier changes. Released sections were corrected in place on the same terms.

[0.7.5]

Changed

• The changelog drops its em dashes. Wording only, with no change to any phase, gate, or trigger phrase; the nine backticked trigger phrases in the 0.6.1 entry are byte-identical. No heading was touched. The released sections corrected in place are 0.7.0, 0.6.1, 0.6.0, 0.4.2, and 0.4.0: their wording changed, their facts did not.
• load-bearing stays in the 0.7.4 entry, because there it is a name. It identifies which paragraph that release removed, and the same name is still live in skills/debug/evals/evals.json and the README. Rewriting it here would leave a reader unable to tell what was removed.
• CHANGELOG.md is declared in scripts/em-dash-purged-paths.txt, alongside the README and skill bodies the gate already defends.

[0.7.4]

Changed

• debug: removed the pre-investigation priming pass and its four bullets, the pressure line at the top of Phase 1, the duplicated load-bearing-artifact paragraph, the duplicated Phase 2 gate, the discovery/glob cost hypothesis, and the inert shell: bash frontmatter key; restated the hypothesis-grounding paragraph once and replaced the Boy Scout sentence with a focused-diff rule that agrees with the skill's own boundary; replaced the description's trigger-phrase list with intent categories; the bundled checklist no longer lets Phase 6 cleanup be skipped for tagged probes, and both bundled files drop their em dashes under the house style
• Applied from the 2026-09 prompt-audit against Claude Fable 5.1 (docs/specs/prompt-audit-skills-2026-09.md).

[0.7.3]

Fixed

• debug: the git pre-compute lines moved out of ## Pre-computed context into a "Repository context. Gather first" body section of individual Bash calls, one command per call, each head bound kept inside its command and a failure read as an unknown value. The harness composes a skill's whole pre-compute block into one shell invocation, and a worktree-isolated session refuses a git-bearing compound command, which blocked these skills from loading inside a worktree. Same shape as the worktree skill's fix in #1619. Non-git pre-compute lines stay where they were.

[0.7.2]

Changed

• Dynamic-context probe fallback made reachable. The working-tree-status injection piped its probe into head before ||, so the fallback could never run and a failed probe rendered an empty string under a label that reads as a clean tree. The fallback now sits in a brace group with the probe and the cap applies outside it. Whole-repo extract-ssot sweep.

[0.7.1]

Changed

• Instruction-surface de-slop (#2891, debugging cluster). Rewrote this plugin's README.md and every SKILL.md to drop em dashes under the repo's zero-tolerance house policy, using /ai-slop:audit fix semantics: periods or commas, or a restructured sentence, never parentheses, en dashes, or a spaced hyphen as a stand-in. Meaning stays; only the mark and the sentence break change.

[0.7.0]

Added

• debug phase 5: the red step now has to be red for the right reason. Step 2 said "Watch it fail (Red)" and stopped there. A test that errors on a typo, a bad import, or an unrelated defect is also red, and the fix that turns that red green has not touched the bug, while the loop reports a clean Red→Green cycle. The step now requires reading the failure message against the root cause being targeted, and repairing the test or the reproduction before any implementation edit when they do not match.

  Absorbed from an upstream cursor/plugins skill (docs/upstream/cursor-pstack.md, the tdd section), and the only part of its seven-step workflow that survived: an adversarial audit of the plan confirmed everything else was owned twice over, and would have let this one slip past unnoticed inside a wholesale rejection. Verified absent by reading the phase before landing.

[0.6.1]

Changed

• /debugging:debug's trigger phrases are now single-quoted. They were written with escaped double quotes inside the double-quoted YAML scalar, and the skill-quality gate's trigger-drop protection tracks only 'single-quoted' phrases, so all nine ('diagnose this', 'debug this', 'why is X broken', 'X is throwing', 'something is wrong with', 'investigate this bug', 'performance regression', 'this is slow', 'intermittent failure') were invisible to it and a future rewrite could have dropped any of them unnoticed. The wording is unchanged; only the quoting is.

[0.6.0]

Removed

• The bare /<skill> alias for this plugin's skills. Their SKILL.md files no longer declare a frontmatter name. The field is optional and defaults to the directory name, so declaring it only restated the path while registering a second, unnamespaced command, which the slash-command picker then echoed back as /plugin:skill (skill). Invoke a skill by its namespaced command; the command itself is unchanged.

[0.5.0]

Added

• debug: standing redaction guard. A new section ahead of the phases requires every secret to be redacted (<REDACTED>) before commands, outputs, or captured artifacts appear in a transcript, work note, or commit; loops read credentials from env vars, captured artifacts are quoted only at the lines carrying the diagnostic signal, and insufficient redacted output routes to the user instead of a wider quote. The no-loop escape hatch now asks for a redacted captured artifact. (Guard from upstream mattpocock/skills diagnosing-bugs v1.2.3; registry: the marketplace repository's docs/upstream/mattpocock-skills.md.)

[0.4.2]

Changed

• Skills with ! dynamic-context injections now declare shell: bash explicitly, per the pinned precompute convention. Bash-only pipelines must not fall through to a PowerShell host.

[0.4.1]

Changed

• Documentation-only: the License section now states the plugin's own MIT license inline and no longer points at a LICENSE file at the repository root, which an installed consumer running from the isolated plugin cache cannot reach. No behavior change.

[0.4.0]

Changed

• BREAKING: /debugging:diagnose is renamed /debugging:debug. The skill runs the full repro → hypothesize → fix → regression-test loop, while "diagnose" promised only the first half and twinned confusingly with /testing:diagnose (a different skill, which keeps its name). Clean break per the marketplace naming effort: no renames-map entry; update invocations to /debugging:debug. "diagnose" stays a trigger word in the skill description. Claude Code's built-in bundled /debug skill is unaffected. The plugin skill has no bare command form and is invoked only as the namespaced /debugging:debug.