The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes content from other skills (SKILL.md files, scripts, and test files) to perform audits and improvements.
Ingestion points: The skill reads skills/<slug>/SKILL.md, associated scripts (e.g., scripts/<slug>.mjs), and test files (test/<slug>.test.ts).
Boundary markers: No specific delimiters or "ignore" instructions are used when interpolating or processing these external file contents.
Capability inventory: The skill uses exec to run gbrain, bun, and jq commands, and has write and edit permissions to modify skill files and test suites.
Sanitization: The instructions do not specify sanitization or validation of the content read from other skills before processing.- [COMMAND_EXECUTION]: The skill frequently uses the exec tool to perform its tasks, including running the gbrain CLI for evaluations, bun test for unit and integration tests, and jq for processing JSON output. This is consistent with its role as a developer utility.

skillify