Use when authoring, packaging or debugging a Mercury skill and its discovery description. Not for session capture via /skillify, agent definitions, hooks or extension packaging.
Scanned 9/22/2026
Install to Claude Code
npx -y skills add Whq02/MercuryCLI --skill skill-forge --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Skill Forge?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/whq02-skill-forge-mercurycli)More formats (shields.io, HTML) on the badges page.
---
name: skill-forge
description: Use when authoring, packaging or debugging a Mercury skill and its discovery description. Not for session capture via /skillify, agent definitions, hooks or extension packaging.
argument-hint: "<skill name or path> [purpose]"
---
# Skill forge
## Locate
- Put `SKILL.md` in a lowercase, hyphenated skill directory.
- Use `<project>/.mercury/skills/<name>/` for project scope, the config home's `skills/<name>/` for user scope, or a declared extension skill directory.
- Account for bundled-name precedence; otherwise the first loaded name wins.
- Keep supporting material in `references/` and offline helpers in `scripts/`.
## Define
- Set `name` to the directory name and supply `description`.
- Start the description with the task and `Use when`; put the exclusion second.
- Keep routing information within the first 250 characters and the whole description within 1,000.
- Add `when-to-use` only for distinct triggers; use `argument-hint` for invocation syntax.
- Set `disable-model-invocation: true` for user-only invocation; set `user-invocable: false` to hide slash completion.
- Use `allowed-tools`, `model` and `effort` only when their overrides are needed; verify model IDs against the live catalogue.
- For skills that deploy, delete, post or pay, require user invocation and describe the effects before execution.
## Write
- Put the procedure first; keep the body within 250 lines.
- Link references at the decision that needs them; do not duplicate their content.
- State how supplied arguments and omitted arguments affect the procedure.
- For disk skills, use `$ARGUMENTS` for invocation arguments and `${MERCURY_SKILL_DIR}` for local paths.
- For bundled skills, resolve helper paths from the supplied base directory; bundled arguments are appended to the body.
- Preserve required steps when cutting an existing skill; remove repetition and unsupported claims.
## Check
- Resolve `${MERCURY_SKILL_DIR}` from the supplied base directory.
- Lint names, frontmatter, description, paths and body size with `python3 "${MERCURY_SKILL_DIR}/scripts/skill_lint.py" <skill-directory>`; use `--self-test` to test the helper.
- Give each helper an offline `--self-test`; run it.
- Invoke the skill on a representative request and confirm the intended body loads.
## Source
Checked: 2026-09-15. [Mercury skill implementation](https://github.com/Whq02/MercuryCLI/tree/fc81e29e4129a56b1bfb3beca9819ebfb29875f9/src/skills).
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!