The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes external content from the repository (READMEs, planning documents, code, and scratch files) to determine which files should be deleted or modified. This creates a surface for indirect prompt injection if those files contain malicious instructions.
Ingestion points: As outlined in steps 2 through 5 of SKILL.md, the agent reads task-related READMEs, interface documentation, existing recovery records, orphaned scripts, and plan/Spec documents.
Boundary markers: There are no specified boundary markers or "ignore" instructions defined to separate the data being analyzed from the agent's core instructions.
Capability inventory: The skill allows the agent to delete files, modify existing documentation (such as lessons and state files), and recommend the execution of other skills like implement or remove-deadcode-py.
Sanitization: The instructions do not specify any sanitization, validation, or filtering of the content found within the ingested files before the agent executes deletions or modifications based on that content.

cleanup