The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, including git diffs, linked issues, and parent specifications.
Ingestion points: The skill reads output from git diff, linked issues, and parent specifications mentioned in the contract.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the processed data are provided to the agent.
Capability inventory: The skill executes shell commands (git fetch, git diff, git status) and suggests performing type-checks and running tests on temporary files.
Sanitization: There is no mention of sanitizing or escaping the content of the diffs or issues before processing.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent to execute shell commands to interact with the version control system and validate code changes.
Evidence: Instructions to run git fetch origin <base>, git diff, and git status to determine the scope of the review.
Evidence: The requirement to "type-check a scratch file" and ensure a "cut test suite still goes red against the old code" implies the execution of external tools like compilers, linters, or test runners (e.g., tsc, jest, pytest).
[PROMPT_INJECTION]: The instructions contain logic that could lead to the degradation of security posture in the target codebase by encouraging the removal of defensive code.
Evidence: The skill defines a "cut" category (ask:) that includes "A guard against a speculative or vanishingly unlikely race or security edge case". This instructs the agent to flag and potentially recommend the removal of security-related code if it perceives the threat as unlikely, which could lead to the introduction of vulnerabilities if the agent's assessment is incorrect.

simplicity-review