jason-hchsieh/probe
skills/probe/SKILL.md
Run a cheap, reversible real probe to ground a belief instead of hallucinating. Use when dreamer confidence is low or when you catch yourself guessing about repository state.
testing
Updated Apr 23, 2026
$ install --global
skillsauthnpx skillsauth add jason-hchsieh/predictive-mind probeInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 24, 2026, 1:10 AM59.8s1 file scanned
SKILL.md
- name:
- probe
- description:
- Run a cheap, reversible real probe to ground a belief instead of hallucinating. Use when dreamer confidence is low or when you catch yourself guessing about repository state.
- allowed-tools:
- Bash(git status:*) Bash(git diff:*) Bash(git log:*) Bash(git branch:*) Bash(ls:*) Bash(pwd:*) Bash(tsc --noEmit:*) Bash(npx tsc --noEmit:*) Bash(pytest --collect-only:*) Bash(pytest -q --collect-only:*) Bash(ruff check:*) Bash(eslint --no-fix:*) Bash(npm ls:*) Bash(node --version:*) Bash(python --version:*) Bash(python3 --version:*) Bash(which:*)
World Models research observation: when the learned dynamics model (M) is unreliable, prefer a cheap real probe over an imagined rollout. This skill is that safety valve.
When to use /probe
- Dreamer confidence < 0.5 for a proposed action.
- You catch yourself guessing about a file's contents, a dependency version, or a test count.
- A PostToolUse hook reported high surprise and you want to re-ground before the next step.
- You are about to state a prediction you can't actually defend.
Cheap, reversible probes (pre-approved)
The following are side-effect-free (or trivially reversible) and are pre-approved in this skill's allowlist:
| Intent | Probe |
|---|---|
| What changed locally? | git status, git diff, git log -5 --oneline |
| What branch am I on? | git branch --show-current |
| What's in this directory? | ls -la <path> |
| Type check without emit | tsc --noEmit |
| What tests exist? | pytest --collect-only -q |
| Are there lint errors? | ruff check <path>, eslint --no-fix <path> |
| Is this dep installed? | npm ls <pkg>, node --version, python --version |
| Is this tool on PATH? | which <cmd> |
Procedure
- State the one-sentence QUESTION you want grounded.
- Choose the MINIMUM probe that answers it. Smaller is better — a probe is not a survey.
- Run it; read the actual output.
- Update your working belief (and, if this probe was prompted by a
prior surprise, consider
/reflectnext).
What /probe is NOT for
- Anything that writes to the filesystem.
- Anything that installs packages or touches the network.
- Anything that takes more than a few seconds.
- "While I'm here, let me also check…" — keep it to the single grounded question.
For everything else, use /predict then run the action normally.
Arguments
$ARGUMENTS — the question you want grounded, in one sentence.
Your response should:
- Restate the question.
- Choose and run the single minimum probe.
- Report the observed answer (ground truth), nothing more.
Related Skills
jason-hchsieh/reflect
tools
VerifiedTrustedCommunity
Surprise-triggered reflection. When a PostToolUse hook reports high prediction error (surprise >= 0.6), invoke /reflect to update beliefs via the belief-reviser and decide whether to continue, replan, reclarify, or ask the user.
SKILL.mdUpdated Apr 23, 2026
jason-hchsieh/predict
tools
VerifiedTrustedCommunity
Record an explicit predicted observation before a side-effecting tool call. Use whenever you are about to run a command, edit a file, or invoke any tool whose outcome you have not already observed. Upholds the predict-before-act invariant.
SKILL.mdUpdated Apr 23, 2026
jason-hchsieh/plan-efe
development
VerifiedTrustedCommunity
Given a goal and a set of candidate next actions, rank them by Expected Free Energy (pragmatic + epistemic value) using past similar episodes, and pick the winner. Delegates scoring to the policy-selector subagent.
SKILL.mdUpdated Apr 23, 2026
jason-hchsieh/checkpoint
testing
VerifiedTrustedCommunity
Snapshot the current belief state (and optionally git-stash the working tree) under a label, so a risky action can be rolled back if it produces high surprise or breaks acceptance tests.
SKILL.mdUpdated Apr 23, 2026