The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from git history and commit messages which could be manipulated to include instructions for the agent.
Ingestion points: The skill runs git log, git diff, and git show in SKILL.md to gather data for analysis.
Boundary markers: No specific delimiters or safety instructions are used to distinguish user code/messages from the skill's own instructions.
Capability inventory: The skill's operations are restricted to reading git data and local reference files; it has no capabilities for network access or persistent file modification.
Sanitization: No sanitization or filtering is applied to the git output before it is processed by the agent.

lesson-learned