The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of external task data.
Ingestion points: The skill ingests untrusted data from storyId (via CLI args, git branch names, or kanban entries), Story Acceptance Criteria (AC), and Technical Notes (Phase 2).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the story content are defined.
Capability inventory: The skill executes shell-based grep commands on the local source code and delegates task operations to worker skills (ln-301, ln-302) with autoApprove=true (Phase 4).
Sanitization: No evidence of sanitization, escaping, or validation of the story content before it is used to build plans or passed to workers.
[COMMAND_EXECUTION]: The skill performs automated discovery by executing commands on the local environment.
Evidence: Phase 2 explicitly instructs the agent to "Grep src/ for error handlers, validators, utilities" to find reusable patterns.
[DATA_EXPOSURE]: The skill accesses multiple sensitive local project files to discover context.
Evidence: It reads docs/tasks/kanban_board.md, CLAUDE.md, and any files matching technical notes or context for the story being processed.

ln-300-task-coordinator