feat: improve skill scores for 5 skills (0% → 92% avg)

[rohan-tessl]

Hey @mcouthon 👋

I ran your skills through tessl skill review at work and found some targeted improvements. Here's the full before/after:

Skill	Before	After	Change
mentor	0%	100%	+100%
consolidate-task	0%	96%	+96%
phase-review	0%	94%	+94%
deep-research	0%	89%	+89%
debug	0%	81%	+81%

All 12 skills scored 0% because allowed-tools was a YAML array ([Read, Grep, Glob]) when the skill spec expects a quoted string ("Read, Grep, Glob"). Since your skills are auto-generated from templates via make, I fixed the templates (source of truth) and regenerated — both Copilot and Claude Code output updated cleanly via make && make validate.

This PR intentionally caps changes to 5 skills so it's easy to review. The included GitHub Action workflow will surface Tessl feedback on future SKILL.md changes, so the rest of the library can improve incrementally.

Changes made

All 5 skills (template + generated):

• Fixed allowed-tools frontmatter from YAML array to quoted string format in cc: section of each template
• Improved description fields with clearer "Use when..." clauses, specific behavior summaries, and additional trigger keywords

phase-review (biggest improvement — 15 → 65 lines):

• Expanded from 7 lines of body to a structured review skill with evaluation criteria (scope, correctness, risk, ordering), output format template, and common-flaws reference table

consolidate-task:

• Enhanced description to clarify the ADR workflow and added trigger keywords ('archive task', 'create ADR', 'task is done')

mentor:

• Rewritten description to lead with user intent and clarify Socratic approach; added 'walk me through this' trigger

deep-research:

• Description now summarizes the actual methodology (breadth-first → depth-first, confidence levels, coverage gaps)

debug:

• Description now outlines the 4-phase process and the 3-attempt circuit breaker; clarified trigger keywords

All changes follow your CONTRIBUTING.md: templates modified, make regenerated output, make validate passes, skills stay under 500 lines.

🤖 Tessl Skill Review GitHub Action

This PR also adds .github/workflows/skill-review.yml so future PRs that touch SKILL.md get automatic review feedback:

• What runs: on PRs that change **/SKILL.md, the tesslio/skill-review action posts one PR comment with Tessl scores and suggestions (updated on new pushes).
• Zero extra accounts: contributors don't need a Tessl login — only the default GITHUB_TOKEN is used.
• Non-blocking by default: feedback-only, no surprise red CI. Add with: fail-threshold: 70 later if you want a hard gate.
• Not a build replacement: this reviews skill markdown only — it doesn't replace your make / generate.js pipeline.
• Why only 5 skills here: keeps this PR reviewable. After merge, every PR that touches SKILL.md gets automatic review, so the remaining 7 skills improve incrementally.

---

Honest disclosure — I work at @tesslio where we build tooling around skills like these. Not a pitch — just saw room for improvement and wanted to contribute.

Want to self-improve your skills? Just point your agent (Claude Code, Codex, etc.) at this Tessl guide and ask it to optimize your skill. Ping me — @rohan-tessl — if you hit any snags.

Thanks in advance 🙏

[mcouthon]

Thanks @rohan-tessl!

A couple of things:

1. I'd like to keep phase-review short as it currently is. It works really well for me, and I don't want to mess with it :)
2. Looking at the docs it seems that the list should be whitespace-delimited, so we should get rid of the commas as well.