molotroka123-cell

skill-forge

Create or improve reusable agent skills from repeated successful workflows and explicit user requirements, keeping them portable across BOSSMAN, OpenCode and Claude-compatible skill directories.

molotroka123-cell 1 Updated 1mo ago
GitHub

Install

npx skillscat add molotroka123-cell/aimaxbossman/skill-forge

Install via the SkillsCat registry.

SKILL.md

Skill Forge

Use when a conversation or completed task reveals a reusable workflow.

Do not create a skill for everything

Create/update a skill when at least one is true:

  • the workflow is likely to repeat
  • the user explicitly asks to remember a procedure as a skill
  • several agents need identical rules
  • a failure exposed a missing reusable checklist
  • a stable tool sequence reduces cost/errors

Do not encode:

  • one-time facts
  • temporary secrets
  • volatile credentials
  • private data that belongs in project memory
  • huge transcripts

Portable location

Preferred canonical location:

.agents/skills/<skill-id>/SKILL.md

OpenCode can discover this location, while BOSSMAN should also scan it.

Optionally mirror/import:

  • .opencode/skills
  • .claude/skills
  • ~/.config/opencode/skills
  • ~/.claude/skills
  • ~/.agents/skills

Do not scan the entire computer. Configure explicit skill source roots.

Frontmatter

Keep portable fields:

  • name
  • description
  • compatibility
  • metadata

Use lowercase kebab-case directory IDs.

Workflow

  1. Identify the repeated goal.
  2. Extract stable rules from the conversation/task.
  3. Separate instructions from facts.
  4. Define triggers: when the agent should use this skill.
  5. Define required tools/permissions in the body or BOSSMAN sidecar metadata.
  6. Define a deterministic workflow.
  7. Define output/acceptance criteria.
  8. Add tests/fixtures when the skill drives code or external actions.
  9. Version the skill.
  10. Record provenance: which task/run motivated the update.

Update policy

Prefer improving an existing skill over creating near-duplicates.

Never silently weaken permission or safety rules when updating a skill.

The Skill Library UI should show:

  • current version
  • diff from previous version
  • agents using it
  • source
  • last successful runs
  • proposed update awaiting approval if it changes permissions