The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local validation command npm run spec:gate <path> to verify that generated specification files comply with the required template structure and mandatory sections.
[INDIRECT_PROMPT_INJECTION]: The skill features a data ingestion surface where it reads from local project configuration and documentation files to inform its output.
Ingestion points: Untrusted data enters the agent context via 02-DOCS/wiki/harness/user-profile.md, 02-DOCS/wiki/sdd/constitution.md, and 02-DOCS/wiki/sdd/config.yaml.
Boundary markers: The skill enforces the use of a strict markdown template (references/spec-template.md) and structural Given/When/Then blocks for acceptance criteria.
Capability inventory: The skill possesses the ability to write files to the 02-DOCS/wiki/sdd/specs/ directory and execute project-local validation scripts.
Sanitization: While explicit sanitization is not detailed, the skill focuses on extracting specific behavioral and technical parameters to guide generation rather than executing content from these files.
[TIME_DELAYED_CONDITIONAL]: The skill implements a versioning mechanism that gates the requirement of certain sections (e.g., 'Cost of not building it') based on the document's timestamp (on or after 2026-09-06). This is a benign conditional check for phasing in documentation standards.
[DYNAMIC_EXECUTION]: The skill utilizes platform-native subagents for automated peer review ('Fresh-eyes spec review') and supports dynamic model routing as configured in the project's SDD settings.

specify