The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill executes a local shell script at .specify/scripts/bash/check-prerequisites.sh. Executing scripts within the repository is a security risk if the script contains malicious logic or if input passed to the script is not correctly sanitized.
Evidence: Workflow Step 1 invokes check-prerequisites.sh with flags like --json and --include-tasks.
Risk: The instructions specifically mention escaping single quotes in arguments (e.g., 'I'''m Groot'), indicating that dynamic or user-provided data is passed directly to the shell, which is a common vector for command injection.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill is designed to ingest and process content from spec.md, plan.md, and tasks.md. Maliciously crafted instructions inside these files could manipulate the resulting analysis report or influence the agent's logic.
Ingestion points: Reads from FEATURE_DIR/spec.md, FEATURE_DIR/plan.md, and FEATURE_DIR/tasks.md.
Boundary markers: No explicit delimiters or 'ignore embedded instructions' warnings are defined for the input content.
Capability inventory: The agent has the ability to execute local bash scripts and generate reasoning-based reports.
Sanitization: None identified beyond manual escaping of single quotes in shell arguments.

speckit-analyze