The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a shell script intended for the agent to execute in the local environment to verify that all requirements (acceptance criteria) are present in the generated implementation plan. The script uses standard utilities such as seq, grep, and sed.
[INDIRECT_PROMPT_INJECTION]: The shell script incorporates a variable (slug) directly into the file path of the command. If the project identifier or filename is influenced by untrusted external data (such as a malicious repository name or specification file), there is a potential surface for command injection within the local shell environment.
Ingestion points: The project slug and the content of docs/<slug>-plan.md are used as inputs to the shell verification process.
Boundary markers: No explicit delimiters or boundary markers are used to isolate the shell command from potential path-based injection characters.
Capability inventory: The skill utilizes shell execution capabilities to read files and perform text searching (grep, sed).
Sanitization: No explicit sanitization or validation of the slug variable is performed before it is interpolated into the shell command string.

walking-skeleton-plan