The Agent Skills Directory

[COMMAND_EXECUTION]: The skill configures a PostToolUse hook in its frontmatter that executes a local Python script (task_status_hook.py) whenever Write, Edit, or Bash tools are used. This is used for maintaining task status.
[COMMAND_EXECUTION]: The instructions guide the agent to execute Python and shell commands to write JSON context files to a session-specific directory for state management.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources such as task files and artifact manifests via MCP tools (sam_read, artifact_read). This data (requirements, constraints, acceptance criteria) influences the agent's subsequent actions during implementation.
Ingestion points: Data enters the context via mcp__plugin_dh_sam__sam_read and mcp__plugin_dh_backlog__artifact_read calls in SKILL.md.
Boundary markers: None explicitly defined in the instructions for the ingested content.
Capability inventory: The skill uses Write, Edit, and Bash tools for implementation.
Sanitization: No specific sanitization or validation of the ingested task content is described.

start-task