Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →xuzhougeng/self-awareness
Wisp-science's actual agent tool surface and runtime boundaries. Load this when deciding which Wisp tool can perform a task, checking whether Python can reach agent or desktop capabilities, choosing between interactive analysis and persisted Runs, or answering questions about delegation, images, skills, memory, artifacts, lineage, credentials, session history, and other self-introspection capabilities.
Passed 9 of 15 test cases with the skill, and 4 of 15 without it.
and 10 more in the full report
9 fixed4 passed either way7 still failing2 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 9 test cases the skill fixed. Same model, same prompt, with and without the skill.
While waiting on `monitor_run` for job `run-4412`, execution paused because `wait_interrupted` returned true. To resolve this, should we resubmit the job with `run_in_context` or cancel and restart it?
| Checked | Without skill | With skill |
|---|---|---|
| Advises against resubmitting the interrupted job. | Missed | Passed |
| Instructs calling monitor_run again with the same run identifier. | Passed | Passed |
The instructions your model reads when the skill loads.
Use only tools advertised in the current conversation. Wisp exposes agent and desktop capabilities as explicit tools; do not assume that an SDK documented by another application also exists here. Some tools are conditional on the desktop session, project settings, execution context, or capability grants. If a tool is not advertised, treat it as unavailable.
Use python for persistent Python analysis and r for persistent R analysis.
Their variables and imports persist per project and execution context. Pass a
context_id such as local, ssh:<alias>, or wsl:<distro> when the work must
run somewhere other than the default local context.
The Python worker initializes an ordinary namespace with common standard-library
modules and any available convenience packages. It does not inject a Wisp
control-plane object. Code executed with python therefore cannot directly call
the agent model, spawn Agents, submit or monitor Runs, inspect Wisp credentials,
or query internal project/session metadata. Leave the Python cell and call the
corresponding Wisp tool instead.
| Need | Wisp interface | Availability and boundary |
|---|---|---|
| Read, create, or patch project files | read, write, edit | Operate on normal filesystem paths within the granted workspace. |
| Find files or text | search, grep | Use before broad manual inspection. |
| Run a short command | shell | Use for bounded foreground commands, not as a long-running job manager. |
| Interactive Python or R analysis | python, r | Persistent per project and execution context; no injected control-plane SDK. |
| Inspect a local image | view_image | Explicit tool call for a supported local image; this is not a Python method. |
| Track a multi-step plan | update_plan | Update task progress when a plan materially helps. |
| Present the completed result | attempt_completion | Wisp's normal completion path; there is no separate structured-output submission SDK. |
| Audit configured workflow guidance | list_skill_catalog | Page through discovered/effective records and use its explicit counts. |
| Discover and load workflow guidance | search_skills, use_skill | Search by task/domain, then load the exact returned skill name. |
| Search confirmed project notes | search_memory | Available only when project memory is enabled. New notes are proposed after a completed turn and require confirmation in the Wisp UI; there is no direct memory-write tool. |
| Delegate multi-file codebase reading | explore | Read-only sub-Agent with its own context and read/grep/search access. |
| Delegate general bounded tasks | delegate_tasks | Desktop-only and capability-gated. Use it only when its schema is advertised; it is not callable from Python. |
| Read a truncated delegated result | get_delegated_result | Desktop-only and available with delegation. Use only when the compact result lacks necessary detail. |
| Submit long-running work | run_in_context | Persist a Run in local, ssh:<alias>, or wsl:<distro>. Prefer this over extending shell timeouts. |
test cases passed, vs 4 / 15 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.