The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill functions by auditing and processing external documentation, code, and configuration files, which creates a surface for indirect prompt injection if the audited files contain malicious instructions.
Ingestion points: The agent reads markdown documents, source code, API contracts, and database schemas from the repository environment.
Boundary markers: The skill does not define specific delimiters or instructions to ignore or isolate potentially malicious data within the files being processed.
Capability inventory: The agent possesses filesystem write permissions, the ability to reorganize folder structures, and can spawn sub-agents to perform parallel tasks.
Sanitization: No explicit sanitization or filtering of the content ingested during the audit is described.
[COMMAND_EXECUTION]: The instructions require the agent to verify that code examples within the documentation are functional, necessitating the execution of arbitrary code found in the repository.
Evidence: The core responsibilities in SKILL.md include an instruction to "Check that code examples in docs actually compile/work", which implies the use of compilers, test runners, or shell execution for untrusted code snippets.

docs-architect