The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the dynamic context injection syntax (!command) in SKILL.md to execute the yaml-helper.sh script at the moment the skill is loaded. This is used to resolve project configuration parameters like automation and qa.level via a relative path from the skill directory.
[DYNAMIC_EXECUTION]: The skill performs dynamic execution by running parse checks, smoke tests, and build commands on source code and test files generated by the agent. It utilizes engine-specific tools (Godot, Unity, Unreal) and npm scripts, using paths determined during the implementation process.
[COMMAND_EXECUTION]: The skill relies on multiple local bash scripts (e.g., story-facts.sh, adr-status.sh, md-section.sh) and project-defined scripts (e.g., npm run smoke) to manage story metadata, verify ADR statuses, and update session checkpoints.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, including story files in production/epics/, architectural registry entries, and decision records, which are interpolated into the brief for the programmer agent.
Ingestion points: Reads story files, TR registry entries, ADR documents, and control manifests from the project directory.
Boundary markers: The skill organizes the handoff brief into distinct sections (Objective, Sources, Boundaries), but does not use unique delimiters or explicit escaping for the ingested content.
Capability inventory: The skill possesses capabilities for file editing, shell command execution, and delegation to sub-agents.
Sanitization: No explicit sanitization or filtering is described for the external story content before it is passed to the programmer agent.

dev-story