The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external ticket files, specifically the 'Acceptance Criteria' and 'Scope' sections, to verify the implementation. This represents a standard indirect prompt injection attack surface where malicious instructions could be embedded in the ticket data.
Ingestion points: Reads ## Acceptance Criteria, ## Scope → In Scope, and ## Scope → Out of Scope sections from project ticket files (e.g., PROJ-42_ticket.md).
Boundary markers: The skill does not explicitly define delimiters or specific 'ignore' instructions for the ticket content, but it requires the agent to 'inspect the actual code' rather than just trust the input.
Capability inventory: The skill can modify source code, execute shell commands (linting, formatting, tests), and perform git operations (git commit).
Sanitization: No explicit sanitization or filtering of the ticket content is described before the agent processes it.
[COMMAND_EXECUTION]: The skill executes standard development tools and version control commands as part of its primary function.
Evidence: The skill uses git commit for version control, git diff for reviews, and linting tools such as ruff format and ruff check. It also executes the project's test suite as defined in the environment.
Context: These commands are routine for a development-focused agent and are subject to the skill's 'Core Rule' of waiting for explicit user confirmation before final execution (git commits).

pavlo-commit-by-commit-execution