The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest, analyze, and summarize the contents of a codebase, which serves as a surface for indirect prompt injection if the project being scanned contains malicious instructions.
Ingestion points: Project manifests (e.g., package.json, pyproject.toml), source code files, route definitions, and git logs as described in SKILL.md and references/discovery-map.md.
Boundary markers: None explicitly mentioned to separate untrusted data from the agent's instructions during the discovery process.
Capability inventory: The skill utilizes filesystem read access, grep for symbol searching, and git commands for history analysis.
Sanitization: The instructions include a specific quality gate in SKILL.md to ensure no environment variable values, secrets, or tokens are recorded, only their names.
[COMMAND_EXECUTION]: The skill utilizes shell-based utilities and git commands to perform repository analysis.
Evidence: In references/discovery-map.md, the skill provides specific shell pipelines using git log, grep, sort, uniq, git ls-files, and xargs to identify code hotspots and large files.

spelunk