Fixed-String Status Responder
A minimal, side-effect-free utility skill used to verify that an agent platform can discover, load, and follow a skill end to end. It performs no file writes, network calls, or command execution; it only instructs the agent to answer with a deterministic token.
When to use
- Smoke-testing a fresh agent installation or a new skill channel.
- Verifying that skill discovery and
SKILL.mdinjection work in a given environment. - CI heartbeat probes where a deterministic, easily-greppable answer is required.
Do not use this skill for real user tasks; it intentionally does nothing useful beyond returning a fixed token.
Parameters
| Parameter | Required | Default | Description |
|---|---|---|---|
token |
no | CANARY_SKILL_LOADED |
The exact string the agent must answer with. |
uppercase |
no | true |
Whether to normalize the token to uppercase before answering. |
Workflow
Step 1: Receive the request
The user (or an automated harness) will phrase the request as a heartbeat / smoke test, for example: "run the canary skill" or "give me the canary token".
Step 2: Compute the token
- Take the
tokenparameter, or the defaultCANARY_SKILL_LOADEDwhen absent. - If
uppercaseis true (default), convert the token to uppercase.
Step 3: Answer
Reply with the token alone on a single line. Do not add explanations, markdown formatting, or surrounding prose. Example response:
CANARY_SKILL_LOADED
Step 4: Stop
After answering, the task is complete. Do not take further actions, do not create files, and do not schedule follow-up work.
Error handling
| Situation | Handling |
|---|---|
| User asks for additional side effects | Politely refuse; explain this skill is a no-op by design. |
| Ambiguous request | Ask the user to confirm they want the canary token. |
Notes for reviewers
This skill exists for security-research and integration-testing purposes. It is deliberately trivial: it contains no scripts, no references, and no executable content.
