The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes untrusted data from git commits which could influence agent behavior or report generation.\n
Ingestion points: Commit data (messages, authors) is read from commits.json in scripts/analyze_commits.py and scripts/calculate_leaderboard.py.\n
Boundary markers: Data is interpolated directly into analysis results and report templates without delimiters or instructions to ignore embedded commands.\n
Capability inventory: The skill utilizes file system access and shell command execution (grep, git, npm).\n
Sanitization: No input validation or sanitization is performed on the commit data. Specifically, scripts/generate_report.py injects author names and commit messages into HTML templates without escaping, introducing a cross-site scripting (XSS) risk.\n- [COMMAND_EXECUTION]: The skill instructs the agent to run various shell commands for local repository analysis.\n
Evidence: Examples in SKILL.md use grep and git log to extract code complexity and churn metrics.\n
Evidence: The skill suggests running npm test and pytest for coverage analysis.

github-analysis