The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data from specification documents and ticket graphs to drive its implementation workflow. This creates a surface where malicious instructions embedded in a ticket could attempt to influence the agent's behavior during the implementation or merging phases.
Ingestion points: The agent is instructed to "Read the full spec and all tickets" in SKILL.md under the 'Read the graph once' workflow step.
Boundary markers: There are no explicit instructions to use delimiters or ignore instructions embedded within the ticket data.
Capability inventory: The skill utilizes shell capabilities for git operations (fetch, switch, worktree) and executes project-specific commands for testing, linting, and building (SKILL.md, steps 3, 4, and 7).
Sanitization: No specific sanitization or escaping of the ticket content is described before it is used to guide the worker's tasks.
[COMMAND_EXECUTION]: The skill relies on executing various shell commands to manage the repository state and verify implementation work.
Evidence: Steps 3 and 4 in SKILL.md detail the use of git fetch, git switch, and the execution of user-defined test, lint, typecheck, and build commands. While standard for development workflows, these commands are executed in the context of the implementation task.

implement-spec