The Agent Skills Directory

[COMMAND_EXECUTION]: The skill functions by executing a series of local commands using the no-mistakes axi utility. These commands control the validation pipeline, including starting runs (axi run), responding to gates (axi respond), and checking status (axi status).
[EXTERNAL_DOWNLOADS]: The skill references external documentation and technical specifications. Specifically, it points to https://kunchenguid.github.io/no-mistakes/ for repository configuration guidance and https://toonformat.dev for the machine-readable output format used by the tool. These resources are maintained by the tool's author and vendor.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it interpolates untrusted user input into the validation pipeline.
Ingestion points: The skill ingests arbitrary text from the user via the /no-mistakes <task> command and the conversation history, which is then passed as the --intent parameter to no-mistakes axi run.
Boundary markers: Absent. The instructions direct the agent to pass the user's task text "verbatim" and "exactly what the user set out to accomplish" into the command line without using delimiters to separate the untrusted text from instructions.
Capability inventory: The skill allows for significant system interaction, including modifying the file system (git commits/branching), executing local subprocesses via the no-mistakes tool, and performing network operations such as pushing to remote targets and managing pull requests.
Sanitization: Absent. The skill explicitly instructs the agent not to condense or drop requirements while capturing the intent, ensuring that any embedded instructions in the user's input are preserved and passed to the downstream validation agents.

no-mistakes