The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill serves as a router that evaluates user requests to delegate tasks to child skills such as strategy, discovery, prds, experiments, and user-stories.
Ingestion points: Processes user requests categorized as strategy, discovery, PRDs, experiments, or user stories in SKILL.md and references/routing-guide.md.
Boundary markers: The skill relies on standard agent behavior; no explicit delimiters or "ignore instructions" warnings are present in the router.
Capability inventory: The router only performs task delegation and chaining to other internal skills; it does not contain direct file system write, network, or subprocess capabilities.
Sanitization: No explicit sanitization or filtering of input before routing is defined.
[COMMAND_EXECUTION]: The README.md file provides installation instructions intended for manual execution by the user. These include the use of npx to add the skill and standard filesystem commands (mkdir, cp) to move skill files into the agent's local directory.
[SAFE]: The .skillmeta.yml configuration includes security-positive path restrictions (forbidden_paths) that explicitly prevent local configuration files, private directories, and user-specific data from being included in upstream contributions, mitigating accidental data exposure.

product-development