The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute shell commands (e.g., pytest, python scripts/reconcile.py) that are defined in a REGRESSION.md ledger file.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository's REGRESSION.md file to determine which commands to run. This presents a surface for indirect injection if an attacker modifies the ledger. However, the skill provides specific instructions to treat these commands as untrusted and requires a human-in-the-loop review process before execution (Step 3 of the Audit Flow).
[SAFE_PRACTICE]: The instructions explicitly warn the agent to treat ledger commands as untrusted repository content and to present the command list and risks to the user for review before execution through a trusted host boundary.

module-regression