The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from project plan files and decision documents, creating a surface for indirect prompt injection attacks.
Ingestion points: The skill accepts a plan file path as an argument and is instructed to read this file, along with decision documents from the file system or git history (SKILL.md).
Boundary markers: The instructions do not define specific boundary markers or delimiters to isolate the content of these documents from the agent's internal reasoning, nor do they provide explicit warnings to disregard instructions embedded within the processed data.
Capability inventory: The skill leverages sensitive capabilities including Bash (for running test suites), Edit, and Write (for modifying files such as moving plans to a 'done' directory) (SKILL.md).
Sanitization: There is no evidence of sanitization, escaping, or structural validation for the text read from external plans or decision documents before the agent evaluates and acts on it.
[COMMAND_EXECUTION]: The skill's primary function involves the execution of shell commands via the Bash tool to verify project state.
Behavior: The agent is instructed to identify and run test commands (e.g., test suites, linters, builds) found within the project environment (SKILL.md).
Git Operations: The skill utilizes git log, git show, git status, and git diff to audit document drift and verify code baselines.

verify-completion