The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use CLI tools like git (blame, log) and gh (pr view) to gather context about code changes.\n- [COMMAND_EXECUTION]: Step 6 specifies the use of other LLM command-line interfaces (codex exec, claude) to cross-reference analysis results using specific safety flags like read-only and plan mode.\n- [DYNAMIC_EXECUTION]: The core logic of the skill involves the agent writing and executing small scripts or tests to verify the safety of a code change. This dynamic script generation and execution occurs at runtime based on the code being analyzed.\n- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted external data from pull requests (titles, bodies, comments, reviews) and third-party library source code.\n
Ingestion points: Output from git blame, git log, gh pr view, and file reads of external libraries in SKILL.md.\n
Boundary markers: No specific boundary markers or instructions to ignore embedded instructions are defined for the ingested data.\n
Capability inventory: The agent can execute shell commands, read/write files, and invoke other model CLIs.\n
Sanitization: No sanitization or validation of the ingested external content is mentioned before processing.

blast-radius