The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/validate_tasks.mjs performs runtime execution of another script (verify_evidence.mjs) by computing its path relative to the project root and invoking it via spawnSync. This pattern represents dynamic loading and execution from computed paths.
[EXTERNAL_DOWNLOADS]: The skill provides instructions in references/specialists.md to download and install additional components from the author's GitHub repository (github.com/promovaweb/specsfy) using the npx skills add command.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from project specification files (spec.md). Content such as task definitions, IDs, and metadata fields are parsed and used to influence the agent's logic, including state transitions and validation outcomes.
Ingestion points: Reads specs/<estado>/<NNNN>-<slug>/spec.md in validate_tasks.mjs and validate_interface_tasks.mjs.
Boundary markers: Relies on specific markdown headings (e.g., 14. Tarefas) and metadata keys, but lacks robust escaping for the interpolated values used in status reports.
Capability inventory: Includes file read/write operations and subprocess execution via node and spawnSync.
Sanitization: Uses regex-based parsing which may be susceptible to malformed inputs designed to bypass validation logic.
[COMMAND_EXECUTION]: The skill instructions and validation scripts frequently invoke external tools and subprocesses (node, npx, spawnSync) to perform context monitoring, database safety checks, and task validation logic.

specsfy-05-tasks