refactor: strip AI slop from authored skills and tests - #4
Merged
Conversation
Drop unused catalog metadata, emoji/CRITICAL padding, stale Make references, leftover refactor test names, and redundant defensive pytest skips. Behavior is unchanged. Co-authored-by: Ben Sykes <SystemFiles@users.noreply.github.com>
Remove the waving-hand PR-title comment, leftover hack/ history in spec_metadata tests, redundant bash skip markers, and track the tree fixture so the git ls-files fallback works without `tree`. Co-authored-by: Ben Sykes <SystemFiles@users.noreply.github.com>
SystemFiles
marked this pull request as ready for review
August 16, 2026 18:00
4 tasks
SystemFiles
pushed a commit
that referenced
this pull request
Aug 19, 2026
* feat(lavish): added a safer version of Lavish AXI skill Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * docs: add AGENTS.md and catalog architecture docs Give agents and contributors a single map of repo layout, Taskfile ops, and contribution rules. Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * feat: add issue-triage skill with scripts, fixtures, and tests (#2) * feat: add issue-triage skill with scripts, fixtures, and tests Ship an explicit-invocation skill that seals GitHub issues with ready/size labels, offline eval fixtures, and script unit tests. Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * fix: address CodeRabbit review on issue-triage Split happy-path eval approval vs seal turns, atomic seal edit, mock_gh repo-scope on edit, SKILL typos, and shlex-quoted GH env. Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.com> --------- Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.com> * feat(issue-triage): persist clarify Q&A with resume and fixed templates (#4) Agents write a resumable clarify log under .issue-triage/, use A–D question templates, and ensure .gitignore ignores that dir when one exists. Adds clarify-resume eval coverage; skill-creator with_skill evals at 100%. Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * chore: remove .lavish directory and add to .gitignore (#6) Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.com> * fix(issue-triage): address multi-model review findings (#5) * fix(issue-triage): address multi-model review findings Correctness (act-on): - seal: cross-check --size against body ## Size section - seal: enforce preflight internally (fail-closed) - validate_sealed_body: enforce section order, non-empty content, adjacent --- before Original Ask, ignore headings inside fences - gh pr list / label list: add --limit 200 to avoid 30-item cap Hardening (consider): - seal: reject XL (sizing heuristics say split, not seal) - preflight: skip refs with missing number instead of crashing - handoff: case-insensitive ready check, sorted size_labels - run_gh: add subprocess timeout (120s) Nits: - SIZE_RE: accept en-dash alongside em-dash/hyphen - _ignored: handle leading /, negation patterns - SKILL.md: align seal checklist order with skeleton - evals.json: sequential ids, add body-file ref to eval #2 - Add clarify-resume/state.json fixture - Remove committed draft-pr/gh-journal.json runtime artifact - Add issue-triage to EXPECTED_SKILLS baseline - mock_gh: enforce repo-mismatch on pr list / label list Tests: 107 pass, 1 skip. task ci clean. Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.com> * fix(issue-triage): search defenced text for SIZE_RE validation SIZE_RE and BODY_SIZE_RE searched the raw body including fenced code blocks. A valid `## Size` line inside a fenced block (e.g. in Original Ask) could mask a malformed real Size section. - validate_sealed_body: search defenced text instead of raw body - issue_ops: defenced body_text before BODY_SIZE_RE cross-check - Add regression test: bad Size + fenced valid Size → fail Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.com> * fix(issue-triage): handle TimeoutExpired, case-insensitive size/ matching - run_gh: catch subprocess.TimeoutExpired → clean stderr + exit 1 - validator subprocess: catch TimeoutExpired → return 1 - cmd_handoff: match size/ prefix case-insensitively, .upper() suffix Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.com> --------- Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.com> * fix(issue-triage): tighten explicit-only invocation and add compatibility (#7) * fix(issue-triage): tighten explicit-only invocation and add compatibility Mirror emerald-grove-pet-clinic-ben-sykes#94 skill conformance: stronger description guardrails, disable auto-trigger prose, compatibility metadata for gh CLI, and a Goal/Intent warning that this skill mutates GitHub. * fix(issue-triage): require named Issue in description triggers Align the frontmatter Use-only-when list with the Goal/Intent guard so every explicit invocation form names a target Issue. Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.com> --------- Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.com> * chore: added upstream matt pocock skills into this library Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * feat: add visual-explain skill for interactive local diff explainers Adapt sighup/claude-workflow cw-explain into an authored skill that writes self-contained HTML under .lavish/ instead of publishing Claude artifacts. Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * chore: vendor mattpocock and obra upstream skills Fix nested catalog paths, sync lock + trees, and document the vendored inventory (grill-*, teach, wayfinder, improve-codebase-architecture, TDD). Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * feat: add sdd-qa skill for one-by-one SDD questions Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * feat: add pr-feedback-qa skill for review disposition Q&A One-item Address/Skip/GitHub Issue flow with Plan mode gate, file or PR input, and resumable .scratch JSON sessions. Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * docs(pr-feedback-qa): drop green/unnumbered batch template Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * docs(sdd-qa): require Plan mode before Q&A Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> * feat: expose catalog as Cursor Agent Plugin (#8) Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> --------- Signed-off-by: Ben Sykes <ben.sykes@liatrio.com> Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Ben Sykes (Liatrio) <bds-liatrio@users.noreply.github.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
2 participants
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.
Why?
Authored skills and tests had leftover AI padding: unused catalog metadata, emoji/CRITICAL emphasis, stale Make references after the Taskfile migration, a no-op directory-exists test, redundant pytest skip markers, and a waving-hand PR-title lint comment.
What Changed?
agentsmd-generator: drop unused frontmatter metadata, emoji intro, and CRITICAL callout; keep the source-of-truth and inventory workflowwork-breakdown: stop pointing at acodebase-explorationskill that is not in this catalogskills/exists check, unused imports, leftover “after refactor” test name, and python3/bash skip noise; point the catalog mismatch hint attask sync-upstream-skills; track the tree fixture sogit ls-filesworks withouttreecat |, shorten a duplicated docstring, fix a stale Makefile commentBehavior is unchanged. Vendored
skills/agent-browser/was left alone.Overlaps #3 (same
/desloprequest). This PR is the more conservative edit: same slop removed, less rewriting of skill instructions.Additional Notes
python3 -m pytest -q→ 74 passed, 1 skipped (jjnot installed)task lint(Task/uvnot in this environment; GitHub Actionstask ciis the gate)SKILL.mdfrontmatter (name+description)task verify-discovery