The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests data from multiple external and project-resident sources without adequate sanitization or boundary enforcement.
Ingestion points: The skill reads various project-level files (build-plan.md, architecture.md, design.md, etc.), the codebase itself, and fetches external content from the web when searching for library documentation.
Boundary markers: There are no explicit instructions or delimiters used to separate untrusted content from the agent's primary instructions.
Capability inventory: The skill has the ability to execute arbitrary shell commands (Bash), modify the filesystem (Write, Edit), and communicate with the user or other agents.
Sanitization: No mechanisms for validating or escaping external data are defined in the instructions.
[COMMAND_EXECUTION]: The skill instructions direct the agent to perform various system-level operations using the Bash tool as part of the software development lifecycle.
It uses git fetch and git rev-list to synchronize with remote repositories and check for changes.
It manages local environments using docker compose up -d and health checks.
It initializes new projects by running framework-specific scaffolding tools and package managers.
It manages database states by generating and applying migrations.
[EXTERNAL_DOWNLOADS]: The skill can fetch content from the internet to assist in development tasks.
It is instructed to spawn subagents with web access to retrieve documentation for libraries and frameworks when current usage is unknown.
[DYNAMIC_EXECUTION]: The skill involves the generation and execution of code at runtime.
Specifically, in the logical track, it generates database migrations and executes them against a live database instance, subsequently verifying the schema changes by performing direct queries.

dev-develop