The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extracts and executes shell commands defined in the 'Deterministic checks' section of specification documents (Phase 2, Step 3). There is no verification, sanitization, or validation of these command strings before they are executed in the host environment.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests and acts upon untrusted data from specification files (typically located at './specs/*.md'). Evidence Chain: 1. Ingestion points: Markdown specification documents and the 'spec_tasks' database table. 2. Boundary markers: Absent; the skill lacks delimiters or warnings to ignore embedded instructions within the parsed content. 3. Capability inventory: Arbitrary shell command execution and the ability to spawn subagents with provided task descriptions. 4. Sanitization: Absent; the skill does not filter or validate extracted commands or task definitions.
[REMOTE_CODE_EXECUTION]: The combination of parsing external Markdown files and executing the contained commands (e.g., 'npm test') creates a vector for remote code execution if the specification files are compromised or generated from malicious prompts.

implement-spec