The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands using the roundfix CLI tool. The commands roundfix spec check <slug> --stage techspec --run-verification and roundfix spec judge <slug> --stage techspec utilize a <slug> parameter obtained from the user-supplied $ARGUMENTS. Direct interpolation of user input into shell commands is a known pattern for command injection vulnerabilities.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant ingestion surface for untrusted or external data that could contain malicious instructions.
Ingestion points: The agent reads the following files to perform its tasks: docs/specs/<slug>/_prd.md, CONTEXT.md, existing ADRs in docs/adr/, and architectural guides in docs/agents/. It also processes 'Inbox Entries', 'Backlog Entries', and 'Findings'.
Boundary markers: There are no instructions provided to wrap these inputs in delimiters or to ignore embedded instructions within these external files.
Capability inventory: The skill possesses the capability to write various markdown files to the local file system (specifications, ADRs, and authorization records) and execute vendor-specific CLI tools (roundfix).
Sanitization: The skill lacks explicit instructions for sanitizing or validating the content of the external files before they are integrated into the agent's reasoning process.

write-techspec