The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external files (SKILL.md, scripts, templates) which could contain malicious instructions designed to influence the agent's behavior during the audit or editing process.
Ingestion points: Reads files from the target repository including SKILL.md, references/**, templates/**, and scripts/** (Section 1).
Boundary markers: The skill does not define specific delimiters or isolation instructions to prevent the agent from following directions found within the files it is analyzing.
Capability inventory: The agent can perform file writes in edit mode (Section 3) and execute shell commands (Section 4).
Sanitization: There is no explicit logic for sanitizing or escaping the content of processed files before they are interpreted by the model.
[COMMAND_EXECUTION]: The skill instructs the agent to run shell commands for validation, specifically npx skills add <repo-root> --list --full-depth and node --test .github/tests/mtm-image2-output.test.mjs (Section 4).
[EXTERNAL_DOWNLOADS]: The use of npx involves downloading and executing packages from the npm registry, which is an external dependency management system.
[SAFE]: The skill references official prompt engineering documentation from OpenAI's developer portal, which is a well-known and trusted service.

gomtm-skills-improve