The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Indirect Prompt Injection risk from git history. The skill ingests commit messages and diffs which are attacker-controllable data sources. Evidence Chain: 1. Ingestion points: SKILL.md Phase 2 (git log and git diff commands). 2. Boundary markers: No delimiters or warnings to ignore instructions within the data are present. 3. Capability inventory: Execution of local git commands. 4. Sanitization: No evidence of input validation or escaping for the ingested content.
[COMMAND_EXECUTION] (SAFE): The skill specifies a set of standard git commands for analysis. While it executes shell commands, they are restricted to non-destructive read operations within the repository context.

lesson-learned