skill-metadata-maintainer
Skill Metadata Maintainer
Maintain metadata in exactly one target skill.
This skill exists to initialize or update the metadata object inside one SKILL.md file while following repository rules from CONTRIBUTING.md.
Non-Negotiable Safety Rules
- You MUST operate on exactly one target skill per request.
- You MUST require the user to specify a unique target, such as the skill name, the skill directory path, or the
SKILL.mdpath. - You MUST NOT infer a bulk target such as "all skills", "the whole repo", or "every missing file" from ambiguous wording.
- If the request does not identify a single skill unambiguously, you MUST stop and ask the user to specify the exact target skill.
- Before modifying anything, you MUST read the existing
SKILL.mdand inspect the currentmetadatavalues if present. - If any target metadata field already has a value and your change would add, replace, or normalize that value, you MUST explicitly show the user what exists now and ask for confirmation before writing.
- You MUST NOT silently overwrite existing metadata values.
Inputs
The request MUST identify one target skill by at least one of:
- skill name, for example
skill-metadata-maintainer - skill directory path, for example
skills/skill-metadata-maintainer SKILL.mdfile path
Optional user intent may also specify:
- initialize missing metadata only
- update selected metadata fields
- normalize metadata to repository convention
Source Of Truth
When generating or validating metadata, follow CONTRIBUTING.md.
Apply these rules:
metadata.name: human-readable display name for the target skill, derived from the skill name by default by removing hyphens and capitalizing each word while preserving well-known brand casing and common all-caps acronymsmetadata.description: purpose statement that describes what the skill is for, not how it works, and should stay under 150 charactersmetadata.author: initialize from the first Git author for the targetSKILL.mdmetadata.created: initialize from the first Git timestamp for the targetSKILL.md, formatted as UTCYYYY-MM-DDTHH:mm:ssZmetadata.version: optional; keep unset by default, and only set or update it when the user explicitly asks for versioning
If Git history for the target SKILL.md is unavailable or incomplete, report that clearly and ask the user how to proceed. Do not invent author or created silently when the intended source of truth cannot be verified.
Required Workflow
Follow this sequence:
- Resolve the target skill to one unique
SKILL.md. - Read the target
SKILL.mdfrontmatter and currentmetadatastate. - Read CONTRIBUTING.md and use it as the governing spec.
- Determine whether the user wants initialization, selective update, or normalization.
- Inspect Git history for the target
SKILL.mdto deriveauthorandcreatedwhen needed. - Determine whether
versionshould remain unset or be set explicitly according to CONTRIBUTING.md. - Prepare the exact metadata object that should exist after the change.
- If any existing field with a current value would be changed, present a concise field-by-field diff and ask for explicit confirmation.
- Only after confirmation, update the target
SKILL.md. - Report the final applied metadata and any assumptions.
Confirmation Policy
You MUST ask for confirmation before writing when any of the following is true:
metadataalready exists- any target metadata key already has a non-empty value
- your update would replace text, not just add a missing field
- your update would normalize formatting for an existing value
- your update would add, change, or remove
version
The confirmation message SHOULD include:
- the target skill identifier
- the exact fields that would change
- the current value and proposed value for each changed field
Output Rules
When no overwrite confirmation is needed, return:
- the resolved target skill
- the metadata values to be written
- the files changed
When overwrite confirmation is needed, return:
- the resolved target skill
- a concise before/after diff for each changed metadata field
- a direct confirmation request before making edits
Red Flags
Stop and ask the user instead of editing if:
- the request names more than one skill
- the path points outside the repository's skill directories
- multiple skills match the provided identifier
- the requested change conflicts with CONTRIBUTING.md
- Git history does not provide a trustworthy source for
authororcreated
Style
- Be strict about target scoping.
- Be explicit about overwrite risk.
- Prefer deterministic edits over inferred broad updates.
- Keep user-facing explanations short and concrete.
More from flc1125/skills
subagent-orchestrator
Orchestrate subagent workflows for complex tasks that benefit from decomposition, role-based delegation, and parallel execution. Use when Codex should assemble a temporary team of subagents, choose roles from a reusable role library, create a controlled fallback role when no preset role fits, coordinate read-heavy work in parallel, or handle write-heavy work with ownership boundaries, staged execution, and an integrator-led merge path.
31yuque
Work with Yuque OpenAPI for reading, searching, creating, and updating users, groups, repos, docs, TOC structures, versions, and statistics. Use when Codex needs to operate on Yuque knowledge bases or documents, reorganize document placement in a repo, inspect API capabilities, or prepare guarded plans for destructive Yuque actions.
30experts
Assemble a panel of experts to assess a problem from multiple professional perspectives, surface agreement and disagreement, and deliver a chaired recommendation with clear tradeoffs. Use when the user wants multi-expert judgment, a second opinion, design critique, option comparison, or a recommendation backed by distinct expert viewpoints.
24study
Guide structured learning for a topic by diagnosing current level, defining stage goals, building a learning path, generating practice, and running review loops. Use when the user wants to learn something step by step, start from zero, build a study plan, prepare for an exam or skill, get guided practice, or continue a topic through staged coaching rather than a one-off answer.
13github-create-pr
Create GitHub pull requests from a local branch using a reviewable workflow for branch checks, diff analysis, PR title/body writing, and gh CLI creation. Use when opening a PR, drafting or improving a PR description, preparing a branch for review, or adding reviewers on GitHub.
13async
Launch and coordinate Codex subagents as deferred tasks. Use when the user wants to start bounded subagent work now, keep the main thread moving without waiting by default, then later join, collect, or redirect that work through a stable task reference.
12