agentleFS
Sign inStart free
ceilf6ceilf6/FrontAgent/skills/skill-lifecycle/SKILL.md

skill-lifecycle

Skill124 starsChanged 5 months agoMITView on GitHub

Create, evaluate, improve, and benchmark content skills using the local Skill Lab workflow. Use when adding a new skill, tuning an existing skill's trigger behavior, iterating on SKILL.md instructions, or deciding whether a candidate skill should replace the current version.

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. 0 are common, 8 are distinctive, and 8 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 8

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when the user wants to work on content skills themselves — creating, evaluating, improving, or benchmarking them.

    line 27
  2. Very specific:

    Read references/workflow.md to choose the right Skill Lab sequence.

    line 38
  3. Very specific:

    Read references/eval-guidelines.md before creating or editing trigger evals.

    line 39
  4. Very specific:

    If the skill does not exist yet, scaffold a new skill package (e.g.

    line 40
  5. Very specific:

    If the skill does not yet have evals, initialize trigger evals (e.g.

    line 41
  6. Very specific:

    If behavior quality matters, initialize behavior evals (e.g.

    line 42
  7. Very specific:

    Run a benchmark before making changes (e.g.

    line 43
  8. Very specific:

    When improvement is requested, generate a candidate and compare it with baseline (e.g.

    line 44
  9. Very specific:

    Only apply a candidate when the benchmark clearly improves and the user wants promotion (e.g.

    line 45
  10. Distinctive:

    Keep the user informed of:

    line 51
  11. Distinctive:

    Prefer benchmark-backed recommendations over intuition.

    line 55
  12. Distinctive:

    Treat the skill package itself as the artifact under iteration:

    line 56
  13. Distinctive:

    Do not trust starter evals blindly.

    line 63
  14. Distinctive:

    Do not auto-apply candidates unless the user requested it or the command explicitly says to do so.

    line 64
  15. Distinctive:

    Do not silently broaden a skill's scope just to improve trigger rate.

    line 65
  16. Distinctive:

    Prefer preserving existing references/assets over inventing new file paths.

    line 66

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Skill Creator

Use this skill when the user wants to work on content skills themselves — creating, evaluating, improving, or benchmarking them.

Trigger

  • Requests to create a new content skill
  • Requests to improve or benchmark an existing skill
  • Requests to reduce false positives or false negatives in skill triggering
  • Requests to compare the current skill against a revised candidate

Workflow

  1. Read references/workflow.md to choose the right Skill Lab sequence.
  2. Read references/eval-guidelines.md before creating or editing trigger evals.
  3. If the skill does not exist yet, scaffold a new skill package (e.g. frontagent skill scaffold <skill-name>).
  4. If the skill does not yet have evals, initialize trigger evals (e.g. frontagent skill init-evals <skill-name>).
  5. If behavior quality matters, initialize behavior evals (e.g. frontagent skill init-behavior-evals <skill-name>).
  6. Run a benchmark before making changes (e.g. frontagent skill benchmark <skill-name>). Use --behavior when behavior evals are available.
  7. When improvement is requested, generate a candidate and compare it with baseline (e.g. frontagent skill improve <skill-name>). Use --behavior to include behavior scoring.
  8. Only apply a candidate when the benchmark clearly improves and the user wants promotion (e.g. frontagent skill promote <skill-name> <candidate-id> or --apply-if-better).

Platform-specific commands listed above use the frontagent skill CLI. See ADAPTATION.md for how to map these steps to a different platform.

Output Contract

  • Keep the user informed of:
    • where eval files live
    • where candidate skills were written
    • whether benchmark scores improved
  • Prefer benchmark-backed recommendations over intuition.
  • Treat the skill package itself as the artifact under iteration:
    • SKILL.md
    • agents/openai.yaml
    • existing references/ and assets/

Guardrails

  • Do not trust starter evals blindly. Encourage editing them toward real prompts before strong conclusions.
  • Do not auto-apply candidates unless the user requested it or the command explicitly says to do so.
  • Do not silently broaden a skill's scope just to improve trigger rate.
  • Prefer preserving existing references/assets over inventing new file paths.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free