The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from architectural decision records (ADRs) and technical preference files to extract rules. While this is the intended purpose, these files represent an ingestion surface for indirect prompt injection if the files are modified by external parties.
Ingestion points: The skill reads multiple markdown files using Glob and Read tools across docs/architecture/ and .claude/docs/.
Boundary markers: No specific boundary markers or delimiters are defined to isolate the content being read from the processing instructions.
Capability inventory: The skill possesses Write permissions to the file system and the ability to spawn sub-tasks.
Sanitization: The skill mitigates risks through a robust protocol involving a manual summary review by the user and a secondary verification gate via the technical-director agent before writing the manifest to disk.
[COMMAND_EXECUTION]: The skill uses allowed tools (Read, Glob, Grep, Write, Task) for documentation management within the repository. No signs of arbitrary or malicious command execution were detected.

create-control-manifest