SkillifySAFE

Teams-first Multi-agent orchestration for Claude Code

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: skillify
description: Turn a repeatable workflow from the current session into a reusable OMC skill draft
---

# Skillify

Use this skill when the current session uncovered a repeatable workflow that should become a reusable OMC skill.

## Goal
Capture a successful multi-step workflow as a concrete skill draft instead of rediscovering it later.

## Quality Gate
Before extracting a skill, all three should be true:
- "Could someone Google this in 5 minutes?" → No.
- "Is this specific to this codebase, project, or workflow?" → Yes.
- "Did this take real debugging, design, or operational effort to discover?" → Yes.

Prefer skills that encode decision-making heuristics, constraints, pitfalls, and verification steps. Avoid generic snippets, boilerplate, or library usage examples that belong in normal documentation.

## Workflow
1. Identify the repeatable task the session accomplished.
2. Extract:
   - inputs
   - ordered steps
   - success criteria
   - constraints / pitfalls
   - verification evidence
   - best target location for the skill
3. Decide whether the workflow belongs as:
   - a repo built-in skill
   - a user/project learned skill
   - documentation only
4. When drafting a learned skill file, output a complete skill file that starts with YAML frontmatter.
   - Never emit plain markdown-only skill files.
   - Do **not** write plain markdown without frontmatter.
   - Minimum frontmatter:
     ```yaml
     ---
     name: <skill-name>
     description: <one-line description>
     triggers:
       - <trigger-1>
       - <trigger-2>
     ---
     ```
   - Write learned/user/project skills to flat file-backed paths:
     - `${CLAUDE_CONFIG_DIR:-~/.claude}/skills/omc-learned/<skill-name>.md`
     - `.omc/skills/<skill-name>.md`
   - Remember that uncommitted skills are still worktree-local until committed or copied to a user-level directory.
5. Draft the rest of the skill file with clear triggers, steps, success criteria, and pitfalls.
6. Point out anything still too fuzzy to encode safely.

## Rules
- Only capture workflows that are actually repeatable.
- Keep the skill practical and scoped.
- Prefer explicit success criteria over vague prose.
- If the workflow still has unresolved branching decisions, note them before drafting.
- Keep `omc-learned` as the storage directory name for compatibility; do not present it as the public invocation name.

## Output
- Proposed skill name
- Target location
- Draft workflow structure or complete skill file
- Verification or quality-gate notes
- Open questions, if any

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.