The Agent Skills Directory

[COMMAND_EXECUTION]: The shell script scripts/preqstation-api.sh uses the eval command within the preq_review_task function to execute shell commands passed as arguments (test_cmd, build_cmd, lint_cmd). This allows for arbitrary command execution and is highly susceptible to shell injection if the arguments contain unsanitized input derived from external task data.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to a user-configured PREQSTATION_API_URL using curl in shell scripts and fetch in the Node.js MCP server. These requests include sensitive authentication headers (Authorization: Bearer) and fetch task data that controls the agent's workflow.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as its core logic relies on fetching task details from a remote API which then dictates the agent's actions.
Ingestion points: Task details, including titles, descriptions, and results, are fetched from the remote API via preq_get_task in both the shell helper and MCP server.
Boundary markers: Absent. The skill instructions do not include delimiters or warnings to ignore instructions embedded in the task data.
Capability inventory: The agent has capabilities to execute shell commands via the preq_review_task helper and perform file system and git operations as part of its implementation workflow.
Sanitization: Absent. External content from task fields is used to determine lifecycle branches and as parameters for verification commands.

preqstation