The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute repository-specific tools (e.g., roundfix) and standard development utilities (e.g., go build, grep, find) as part of the preflight and verification stages of task decomposition.
[DYNAMIC_EXECUTION]: The skill generates Markdown task files containing shell commands in a ## Verification section. These commands are subsequently executed using roundfix spec check <slug> --run-verification to validate that the tasks are correctly defined and functional.
[INDIRECT_PROMPT_INJECTION]: The skill derives task definitions and verification logic from untrusted external documents (_prd.md and _techspec.md). A maliciously crafted specification could influence the generation of shell commands executed during the verification process.
Ingestion points: Processes content from docs/specs/<slug>/_prd.md and docs/specs/<slug>/_techspec.md.
Boundary markers: Generated artifacts are organized within the docs/specs/<slug>/ directory.
Capability inventory: The skill possesses the ability to execute shell commands via the roundfix tool and compiler toolchains.
Sanitization: There is no evidence of content sanitization or validation of the input specifications before they are used to generate executable verification commands.

write-tasks