The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from the conversation history and the codebase to generate a specification. While there is a risk that malicious instructions in the code or conversation could influence the final spec, the impact is limited to the content of the generated issue.
Ingestion points: The skill reads from the current conversation context and codebase understanding as defined in SKILL.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the spec template.
Capability inventory: The skill instructs the agent to "publish it to the project issue tracker," which implies write access to an external service (likely via a command-line tool like gh).
Sanitization: No explicit sanitization or validation of the synthesized content is mentioned before publishing.
[COMMAND_EXECUTION]: The skill requires the agent to "Explore the repo" to understand the codebase. This involves file system traversal and reading, which is standard for an engineering assistant skill but should be monitored for scope.
[DATA_FLOW]: The skill moves data from the local development context (codebase/conversation) to an external issue tracker. This is the primary intended function of the skill and is documented in the description.

to-spec