The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the bash tool to execute npx skills commands for managing skill lifecycles, including listing, adding, and updating skills.
Evidence: The skill instructs the agent to run npx skills list --global --json, npx skills update <skill...> --global --yes, and npx skills add <published-source> --skill <skill...> --global --agent '*' --yes.
[EXTERNAL_DOWNLOADS]: The skill uses npx to execute the skills package, which may result in downloading the package from the npm registry if it is not already present in the environment.
Evidence: Multiple shell blocks utilize npx skills without specific version constraints.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from repository files and metadata to drive its publication logic, creating a surface for indirect prompt injection.
Ingestion points: Reads SKILL.md (frontmatter), repository instructions, branch status, manifest conventions, tests, CI logs, and release documentation.
Boundary markers: Instructions mandate resolving the exact primary target from evidence and require confirmation before remote writes, merges, or publications.
Capability inventory: The skill is granted Read, Write, and Bash tools, which allow for file modification and arbitrary command execution.
Sanitization: The skill relies on external validators and repository-defined tests, but lacks explicit internal sanitization or escaping of the ingested content before it enters the agent's context.

publish-agent-skill