The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to perform filesystem management using shell commands, specifically mkdir -p for directory creation and git mv for moving source documents into specification folders. It also explicitly calls for the execution of the roundfix CLI tool (e.g., roundfix spec check and roundfix spec judge) to verify the generated artifacts.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources, which presents a risk of indirect prompt injection where malicious instructions embedded in those sources could influence agent behavior.
Ingestion points: The agent reads from inbox/<repository>/, docs/findings/, and docs/backlog/, and is also instructed to perform web research for market and competitor context.
Boundary markers: While the skill enforces a strict PRD template and glossary declaration process, it does not specify the use of clear delimiters or instructions to ignore embedded commands within the ingested source material.
Capability inventory: The skill has the capability to write to the filesystem (creating _prd.md, _index.md, and _authorization.md), move repository files via git, and execute external CLI tools.
Sanitization: No explicit sanitization, filtering, or escaping of the content retrieved from inbox entries or findings is mentioned before it is incorporated into the PRD.
[DYNAMIC_EXECUTION]: The skill involves the dynamic execution of the roundfix tool suite to perform automated checks and advisory judgments on the documentation produced at runtime. Notably, the skill includes a safety instruction advising the agent to never print, store, or ask for API keys (e.g., ROUNDFIX_OPENROUTER_API_KEY) associated with the tool's backend services.

write-prd