The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests instructions from external task and ticket files that directly influence its execution flow and the commands it runs.
Ingestion points: SKILL.md (Step 1) and execute.md (Step 1) describe reading ticket files, task files, and design.md to determine the work checklist and validation steps.
Boundary markers: The skill defines a "ticket footprint" in SKILL.md to restrict file write operations, but it lacks specific boundary markers or instructions to ignore malicious commands embedded within the data files it processes.
Capability inventory: The skill possesses the capability to execute shell commands (git operations and validation commands) and to spawn/dispatch sub-agents with delegated authority.
Sanitization: There is no evidence of sanitization, filtering, or validation performed on the commands extracted from the "Validation" section of the markdown task files before execution.
[COMMAND_EXECUTION]: The skill is designed to run arbitrary shell commands defined in the "Validation" section of task files (execute.md, Step 4). This allows the content of data files to dictate executable actions on the system.
[DYNAMIC_EXECUTION]: The skill dynamically extracts and executes shell commands provided in external markdown documentation at runtime, effectively treating data as code to verify task completion.

implement-task