develop-skills
Author, audit, merge, and evaluate Agent Skills and instruction stacks. Explicit invocation only.
How do I install this agent skill?
npx skills add https://github.com/sjunepark/agent-scripts --skill develop-skillsIs this agent skill safe to install?
- Gen Agent Trust Hubpass
This skill provides developer workflows for creating, auditing, and merging Agent Skills. It is generally safe and adheres to high-quality documentation standards, but it possesses an inherent attack surface for indirect prompt injection because it is designed to ingest and process untrusted skill files while maintaining access to local shell commands for validation and integrity checks.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
Develop Skills
Produce a focused skill or instruction change that improves the requested work. Use task evidence and concrete constraints; retain only guidance that changes a decision, supplies non-obvious knowledge, or makes completion verifiable.
Choose the relevant guidance
Read the route needed for the requested operation, not every workflow:
- Create or revise for skill authoring.
- Audit instructions for instruction-stack or catalog audits, including model migrations.
- Merge skills for consolidation and predecessor retirement.
- Evaluate a skill for validation and behavioral or selection evidence. For edits, use the checks relevant to what changed.
- Authoring rubric for reviewing the candidate's scope, instruction quality, and completion boundaries.
- Portability contract when creating a package or changing its structure, metadata, dependencies, or cross-client assumptions.
- Codex invocation policy when creating or changing Codex adapter settings.
- Refresh upstream guidance only for an explicit request to reconcile this skill with authoring sources.
Constraints and completion
Preserve existing invocation policy unless a change is requested. This catalog's authoring preference is explicit invocation for new skills; implicit discovery needs evidence of recurring use, reliable matching, and useful activation within the intended installation scope. Installation reach is a separate choice.
Honor the user's operation and existing authorization. A read-only audit stays read-only; a request to revise authorizes necessary in-scope source edits. Publication, installation, and removal of active predecessors must be covered by the requested transition. Resolve only missing authority, not permissions already given, and continue independent work.
Keep runtime guidance independent of incidental host or model assumptions. Skills targeting a named tool may declare its actual prerequisites. Preserve licensing and attribution for retained third-party material.
Finish the authorized change through validation and fixes, without stopping at a first draft. Report changed or deliberately retained behavior, checks and observations, and concrete limits. Static checks, simulated decisions, and live execution support different claims; identify which evidence you obtained.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/sjunepark/agent-scripts/develop-skills">View develop-skills on skillZs</a>