The Agent Skills Directory

[COMMAND_EXECUTION]: The bundled script scripts/inspect_pr_checks.py invokes git and gh using subprocess.run. These calls use argument lists rather than shell strings, which is the recommended secure method for executing external processes to prevent command injection. This functionality is essential for the skill's purpose of inspecting repository CI status.\n- [PROMPT_INJECTION]: The skill processes GitHub Actions logs, which are external content that can serve as an indirect prompt injection surface. This is a common and necessary risk for CI debugging tools, which is mitigated here by requiring explicit user confirmation before taking any action based on the agent's analysis.\n
Ingestion points: GitHub Actions logs are fetched via the gh tool and processed in scripts/inspect_pr_checks.py.\n
Boundary markers: Log snippets are presented to the agent without specific isolation tokens or 'ignore embedded instructions' warnings in the provided script output.\n
Capability inventory: The skill can execute local scripts, standard CLI tools (gh, git), and is instructed to propose or apply code modifications to the local repository.\n
Sanitization: Log content is extracted and displayed as raw text without specific filtering for malicious instruction patterns.

gh-fix-ci