| name | probe-loading |
| description | Benchmark skill for testing loading timing, resource enumeration, content presentation, and lifecycle behavior. Use when asked to probe loading behavior. |
| allowed-tools | Read Write |
| compatibility | Requires filesystem access |
| metadata | {"benchmark":"true","version":"1.0"} |
Loading Behavior Probe
This skill exists to test how platforms load skill content. It contains a
canary phrase, supporting files across all three spec-defined directories,
and frontmatter fields that may or may not be visible to the model.
Canary Phrase
The canary phrase for this skill is: CARDINAL-ZEBRA-7742
If a tester asks you whether you know the phrase CARDINAL-ZEBRA-7742 without
explicitly activating this skill, and you can answer, it means the platform
loaded this skill's body at discovery time rather than waiting for activation.
Instructions
When this skill is activated, respond with a structured report:
-
Frontmatter visibility: Can you see the YAML frontmatter above these
instructions? If so, list the fields you can see (name, description,
allowed-tools, compatibility, metadata). If not, note that frontmatter
was stripped.
-
Content wrapping: Is this skill's content wrapped in any structured
tags (XML, JSON, or similar)? If so, describe the wrapping format.
-
Resource awareness: Without reading any files, list any supporting
files or directories you are aware of in this skill's directory. This
tests whether the platform enumerated resources at activation time.
-
Resource content: Without reading any files yourself, do you already
have the contents of any files from this skill's references/,
scripts/, or assets/ directories in your context? If so, list which
files' contents you can see. This tests whether the platform eagerly
loaded resource content.
References
For detailed API documentation, see API reference.
For error codes, see error reference.
