The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests specifications (spec.md) and project source files (e.g., React components, Laravel controllers) to drive the implementation of tasks and the execution of tests and migrations. This creates an attack surface where malicious instructions embedded in project data or specs could influence agent behavior.
Ingestion points: specs/<estado>/<NNNN>-<slug>/spec.md (containing tasks, Acceptance Criteria, and Functional Requirements) and the project's own source code files.
Boundary markers: The skill does not explicitly use delimiters or "ignore embedded instructions" prompts when reading the specification or codebase content.
Capability inventory: The skill can execute shell commands (artisan, prisma, node), write to the filesystem, perform database migrations, and perform automated skill transitions (e.g., to $specsfy-documentator) without manual user confirmation.
Sanitization: verify_evidence.mjs includes path validation logic to ensure that file access and hashing operations remain within the project's root directory.
[EXTERNAL_DOWNLOADS]: The skill includes instructions to download and install extension skills ("specialists") from the vendor's GitHub repository (github.com/promovaweb/specsfy) using npx. These are vendor-owned resources used for project-specific expertise (e.g., Laravel, Docker).
[COMMAND_EXECUTION]: The skill executes several local Node.js utility scripts included in the package (next_task.mjs, verify_evidence.mjs, render_delivery.mjs) to manage the implementation lifecycle, validate evidence, and select the next task. It also invokes project-level CLI tools for testing and database management (like Laravel's Artisan, Prisma, or Supabase) based on the project configuration.

specsfy-07-implement