The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill's core workflow requires the AI agent to dynamically generate and execute scripts on the local system to verify software hypotheses.
Evidence: 'Step 5: Prove the one fact... Write the script or test that runs the real code, run it, and paste what happened.'
Evidence: 'Rung 4 is usually one small script that imports the same module production runs and calls the exact function you are worried about.'
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from a codebase, which serves as the foundation for the scripts it eventually executes.
Ingestion points: Source code files, 'git log' output, system prompts, agent files, and configuration schemas (SKILL.md).
Boundary markers: Absent. The instructions do not specify the use of delimiters or warnings to ignore instructions embedded within the code being analyzed.
Capability inventory: The agent is instructed to use its execution environment to run scripts that can import production modules, perform network operations, or modify files.
Sanitization: Absent. There are no instructions for sanitizing or validating the symbols, field names, or logic extracted from the codebase before they are interpolated into the generated scripts.
[COMMAND_EXECUTION]: The skill requires the agent to perform extensive shell-based investigations of the repository history and structure.
Evidence: Use of git log -S, git log -L, and grep commands to trace symbols and function history.

what-could-break