The Agent Skills Directory

[DYNAMIC_EXECUTION]: The files scripts/generate_report.py and scripts/create_github_comment.py contain a generate_script method that constructs a shell script from the template parameter and applies chmod 0o755 to the resulting file. This allows the creation of executable files from arbitrary strings.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from CI/CD pipeline outputs and build logs.
Ingestion points: scripts/analyze_metrics.py reads data from a target directory, and SKILL.md instructions include collecting performance metrics from a build.
Boundary markers: None identified.
Capability inventory: allowed-tools include Bash (ci, metrics, testing namespaces), Write, and Edit. Scripts can write and modify executable files.
Sanitization: None; the content is directly interpolated into a shell script template.
[COMMAND_EXECUTION]: The skill's configuration in SKILL.md allows the use of Bash across multiple scopes, which increases the potential impact of the script generation functionality if misused.

detecting-performance-regressions