The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the project's codebase and a ledger file during the verification process, creating a surface for potential instruction injection.\n
Ingestion points: The agent reads the deferred-work.md ledger (in SKILL.md Step 1), deferred-work-archive.md, and arbitrary files/components within the codebase (in SKILL.md Step 2) as specified by the entries it is triaging.\n
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are utilized when processing the content of the ledger or code during the verification phase.\n
Capability inventory: The agent can execute shell commands (grep, git log, echo), read arbitrary files in the repository, and write to specific JSON result files or the ledger file during migration mode.\n
Sanitization: The instructions do not provide requirements for sanitizing or escaping the content read from external codebase files or ledger entries before they are used to influence the triage categories.\n- [COMMAND_EXECUTION]: The skill instructs the agent to run local shell commands for environment validation and code inspection.\n
Commands like grep and git log are used to search for symptoms of issues in the source code to verify the status of ledger entries.\n
Environment checks are performed using echo to verify the execution mode and locate the ledger file path.

bmad-loop-sweep