The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from git diffs and GitHub Pull Requests, which creates a surface for indirect prompt injection.
Ingestion points: Data enters the system via git diff and gh pr diff output as described in SKILL.md.
Boundary markers: Instructions mandate an "evidence-first" contract and require the agent to cite specific file paths and line ranges, which helps prevent the agent from following instructions embedded in the code being reviewed.
Capability inventory: The skill can execute shell commands (git, gh, python) and launch a browser via the open command.
Sanitization: The generate_html_report.py script applies HTML escaping to all content derived from external sources before including it in the report.
[COMMAND_EXECUTION]: The skill utilizes shell commands to perform its core functions, such as invoking git and gh to extract diffs, running local Python scripts for analysis, and using the open command to display reports.
[EXTERNAL_DOWNLOADS]: The generated reports reference external assets for functionality, such as fetching highlight.js scripts and styles from Cloudflare's public CDN to enable syntax highlighting.

code-review