The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests and processes data from external sources, such as user-provided ticket URLs and specification files, alongside the local codebase. This creates a potential surface for indirect prompt injection if those external sources contain malicious instructions.
Ingestion points: External references (spec paths, ticket numbers, or URLs) and local codebase files as described in the 'Gather context' and 'Explore the codebase' steps.
Boundary markers: No specific boundary markers or 'ignore' instructions are defined to isolate untrusted external content.
Capability inventory: The skill possesses the ability to read codebase files, write local files (in the .scratch directory), and interact with external ticket tracking platforms.
Sanitization: The instructions do not specify any sanitization or validation logic for the content fetched from external references.
[DATA_EXFILTRATION]: The skill is designed to read codebase information and transmit summarized data to external trackers like GitHub or Linear, which constitutes a legitimate but functional exfiltration path.
[EXTERNAL_DOWNLOADS]: The agent is instructed to fetch and read the full content of external URLs provided by the user during the context gathering phase.
[COMMAND_EXECUTION]: The skill directs the agent to 'explore the codebase,' which involves executing file system navigation and read commands to understand the project structure.

to-tickets