The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run various shell commands to verify project state, including project test runners, build commands, and linters. It specifically uses a shell pipeline involving git ls-files, sort, xargs, and sha1sum (or shasum) to generate a content hash of the working tree.
[COMMAND_EXECUTION]: The skill executes external scripts or 'recipes' located at .pi/skills/verify-<app>/ to perform end-to-end testing of application features. These recipes can include starting/stopping the application and driving its interface.
[INDIRECT_PROMPT_INJECTION]: The agent is instructed to read and interpret the output of executed commands (e.g., test results, build logs) to confirm if criteria are met. If a project contains malicious code designed to produce deceptive output, it could attempt to influence the agent's subsequent actions or claims.
Ingestion points: Command output files located in /tmp/verify-<hash>-<name>.log and project files read during hashing.
Boundary markers: None identified in the instructions for command output processing.
Capability inventory: Subprocess execution (tests, build, linter, git, hashing tools), file read access (project source), and file write access (logging to /tmp).
Sanitization: No explicit sanitization or filtering of command output is mentioned before interpretation.

verify