eval-audit
**Locale**: All templates in this spec are written in English. Detect the user's language from the session and translate user-facing text at display time per…
**Locale**: All templates in this spec are written in English. Detect the user's language from the session and translate user-facing text at display time per SKILL.md's Global UX Rules. Dimension labels: see the canonical table in SKILL.md.
$ npx -y skills add Evol-ai/SkillCompass --agent claude-codeHow it fires
How this command gets triggered: by you, by Claude, or both.
/skill-updateContext preview
What this command does when you run it.
**Locale**: All templates in this spec are written in English. Detect the user's language from the session and translate user-facing text at display time per SKILL.md's Global UX Rules. Dimension labels: see the canonical table in SKILL.md.
> **Locale**: All templates in this spec are written in English. Detect the user's language from the session and translate user-facing text at display time per SKILL.md's Global UX Rules. Dimension labels: see the canonical table in SKILL.md.
Check installed skills for remote updates and guide the user through updating and evaluating them.
Use `lib/update-checker.js` to scan all skills:
const { UpdateChecker } = require('./lib/update-checker');
const checker = new UpdateChecker();
const stale = checker.getStale(inventory, 7); // not checked in last 7 days
const allGit = checker.checkAll(inventory); // all git-repo-backed skillsExecute via **Bash** `node -e`.
If no git-based skills found:
No git-repo-backed skills found; cannot check for updates. For manually installed skills, check the source page for new versions.
Stop.
**If called with `all`**: skip selection, check all git skills.
**If called with `<skill-name>`**: skip selection, check that skill only.
**If called with no args**: show the list and let user choose. Follow gstack's AskUserQuestion pattern — re-ground context, simplify, recommend, show options with effort scale.
{N} skills available to check for updates (git repo-backed):
1. code-review last checked: 12 days ago github.com/user/code-review
2. superpowers last checked: 9 days ago github.com/obra/superpowers
3. doc-writer last checked: 15 days ago github.com/user/doc-writer
4. ecc last checked: 8 days ago github.com/affaan-m/everything-claude-code
Checking updates requires network (git fetch), about 5 seconds each.
Enter a number, a name, or comma-separated list (e.g. 1,3), or type "all".
You can also type any other skill name directly.
[Check all / Select / Skip]User can respond:
For each selected skill, run `checker.fetchAndCheck(skillPath)` via **Bash** `node -e`. Show results one by one:
**Has updates:**
Checking code-review... ✓ fetch done · 2 new commits found (v1.2.0 → v1.3.0) [Update and evaluate (recommended) / View changes / Skip]
If user chooses "View changes", run `git log HEAD..FETCH_HEAD --oneline` and show, then re-offer the choice.
**Already up to date:**
Checking superpowers... ✓ Already up to date
**Fetch failed (network error):**
Checking doc-writer... ⚠ Cannot reach remote (network issue or repo removed)
When user chooses "Update and evaluate":
1. **Snapshot before pull**: Ensure the current version has a snapshot file in the version management system, so `/eval-rollback` can restore it if the update causes regressions.
a. Use the **Read** tool to load `.skill-compass/{skill-name}/manifest.json`.
b. **If manifest doesn't exist**: create one per `shared/version-management.md` § Creating manifest (version = current upstream version extracted from SKILL.md frontmatter, or `1.0.0`; trigger = `"initial"`). Save the current SKILL.md to `.skill-compass/{skill-name}/snapshots/{version}.md`.
c. **If manifest exists**: read `current_version` from it. Check whether `.skill-compass/{skill-name}/snapshots/{current_version}.md` exists. If not, save the current SKILL.md there. No new manifest entry needed — the version is already tracked; we're just ensuring its snapshot file is present.
d. Pull:
const { UpdateChecker } = require('./lib/update-checker');
const checker = new UpdateChecker();
const result = checker.pullUpdate(skillPath);2. If pull fails — dirty tree:
⚠ Working tree has uncommitted changes. Commit or stash local edits before updating. [View changes / Handle manually]
3. If pull fails — not fast-forwardable (diverged history):
⚠ Update conflict — local branch has independent commits; cannot auto-merge. [View diff / Abandon update / Handle manually]
4. If pull succeeds:
(Note: `git pull` is a shell operation, not a Write/Edit tool call, so `eval-gate.js` does NOT auto-trigger.)
✓ code-review updated to v1.3.0
D1=9 D2=8 D3=9 ✓ Quick scan passed
Last full eval: 82 points (v1.2.0). Version changed; recommend re-evaluating to confirm quality.
[Full evaluation (recommended) / Skip]5. If user chooses "Full evaluation":
After all selected skills are checked:
✓ Update check complete
Checked {N} · updated available {U} · updated {A} · skipped {S}
[Back to /skillcompass / Done]Evaluate agent skill quality. Find the weakest link. Fix it. Prove it worked.
Repo: Evol-ai/SkillCompass
**Locale**: All templates in this spec are written in English. Detect the user's language from the session and translate user-facing text at display time per…
**Locale**: All templates in this spec are written in English. Detect the user's language from the session and translate user-facing text at display time per…
**Locale**: All templates in this spec are written in English. Detect the user's language from the session and translate user-facing text at display time per…
- **Recommended model: Claude Opus 4.6** (`claude-opus-4-6`). Directed improvement requires understanding complex rubric feedback and generating precise,…
**Locale**: All templates in this spec are written in English. Detect the user's language from the session and translate user-facing text at display time per…
**Locale**: All templates in this spec are written in English. Detect the user's language from the session and translate user-facing text at display time per…