Index of every page on this site: llms.txt
All skills
hkuds avatar

/skill-evolution

@dd38286
Run it yourself

Evolve new skills from project experience. Use when a repeating pattern, useful workflow, or hard-won knowledge should be captured as a reusable skill for the current project.

  • 1 file
  • 2.5 KB
  • Updated last month
  • GitHub
Use this Skill: https://skilld.dev/gh/hkuds/autoswarm/skill-evolution

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈48 tokens for metadata: the name and description. ≈586 when used: this file.

Description uses 2.4% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈193 characters, or 2.4%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Skill Evolution

Distill project experience into reusable skills that live under the current project.

Two evolution paths

  1. Automatic: The system promotes playbook skills when an employee's pattern reaches 2 project reflections with recurring behaviors, checklists, or preferences. These are saved as <employee>-<role>-<domain>-playbook and require no manual action.
  2. Agent-driven (this skill): You proactively identify valuable patterns and create skills manually. Use this when domain knowledge, workflows, or tool sequences should be preserved before the auto-promotion threshold is reached, or when the knowledge is not tied to a specific employee pattern.

When to evolve a skill (agent-driven)

  • A task required non-obvious steps that will likely recur
  • You discovered a workflow that worked well and should be preserved
  • Domain-specific knowledge was gathered that future tasks will need
  • A sequence of tool calls forms a reliable pattern worth codifying

Process

  1. Identify the pattern: What knowledge or workflow is worth preserving?
  2. Read the skill-creator skill: file_read the skill-creator/SKILL.md for format and naming guidelines
  3. Name the skill: Lowercase letters, digits, and hyphens only, under 64 characters (e.g., api-validation-workflow)
  4. Create the skill directory: Under .opc/projects/<project_id>/skills/<skill-name>/
  5. Write SKILL.md with proper frontmatter (name, description) and concise instructions
  6. Add scripts/references/assets if the skill benefits from bundled resources

Skill location

Evolved skills belong to the project that produced them:

.opc/projects/<project_id>/skills/<skill-name>/SKILL.md

Guidelines

  • Keep it concise: only include what the agent doesn't already know
  • Use concrete examples over abstract explanations
  • Include the minimal set of steps needed to reproduce the workflow
  • Add scripts/ for deterministic operations that get rewritten repeatedly
  • Add references/ for domain docs the agent should consult
  • Follow the same naming conventions as auto-promoted skills (hyphen-case)

Cross-project reuse

Other projects can reference this project's skills via file_read when the secretary recommends them. Do not duplicate skills across projects; read from the source project instead.

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at dd38286. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 10 hours ago.

Activeupdated last month

README badge

README badge for hkuds/autoswarm/skill-evolution