The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources such as PRD and ADR files to generate implementation plans. This creates a surface for indirect prompt injection where malicious instructions embedded in these source documents could influence the agent's planning output or downstream coding tasks.\n
Ingestion points: The skill reads content from docs/prds/, docs/adr/, docs/intent/, and project files using the Read, Glob, and Grep tools.\n
Boundary markers: The instructions do not define the use of explicit delimiters or boundary markers when processing these external inputs.\n
Capability inventory: The skill possesses the ability to Write plan files to the filesystem and invoke other skills like humanizer.\n
Sanitization: The instructions specifically mandate that acceptance criteria and glossary terms be copied "verbatim" from the PRD, which ensures that any malicious prompt content in those sections is preserved and processed by the agent.

blueprint