AgentStack
Sign in Publish
MCP servers Skills Toolkits Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Observe

skill-ghostlygawd-recursive-harness-observe · by GhostlyGawd

Record falsifiable predictions, score outcomes, inspect calibration and a compact scorecard, or audit/delete Recursive Observe's private local evidence. Use when a user asks to track whether an agent task succeeds, measure confidence, review prediction accuracy, or manage Observe data without changing the current repository.

↗ Repository
— No reviews yet
0 installs
1 views
0.0% view→install

Install

$ agentstack add skill-ghostlygawd-recursive-harness-observe

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • ✓ Prompt-injection patterns
  • ✓ Secret / credential exfiltration
  • ✓ Dangerous shell & filesystem operations
  • ✓ Untrusted network calls
  • ✓ Known-malicious package signatures

What it can access

  • ✓ Network access No
  • ✓ Filesystem access No
  • ✓ Shell / process execution No
  • ✓ Environment & secrets No
  • ✓ Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Observe? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Observe

Use the bundled deterministic CLI. Do not emulate its ledger with prose or create project files. Resolve scripts/observe.py relative to this SKILL.md; never run an untrusted project-local file with the same name.

Record and score

Before uncertain, meaningful work, record one falsifiable expected result:

python3 /scripts/observe.py predict \
  --task "harden the parser" \
  --expect "the malformed fixture is rejected and the full suite stays green" \
  --confidence 0.75

After observable evidence exists, score the printed identifier. Never infer success from intent or score an unfinished task.

python3 /scripts/observe.py outcome PREDICTION_ID \
  --result hit --notes "fixture and suite passed"

Keep task, expect, and notes short. Do not include prompts, source contents, secrets, credentials, or personal data. The runtime applies defense-in-depth redaction before writes.

Review evidence

Use stats for calibration buckets or scorecard for the compact proof surface. Add --json when another tool needs structured output.

python3 /scripts/observe.py stats
python3 /scripts/observe.py scorecard --json

Protect privacy

Observe stores sanitized evidence below ~/.recursive-harness/observe, never the working repository. The runtime intentionally accepts no state-path argument or environment override. Read [the privacy contract](references/privacy.md) when the user asks what is stored, where it lives, how long it remains, or how uninstall affects it.

Audit aggregate metadata without printing prediction text:

python3 /scripts/observe.py privacy audit --json

Delete all Observe records only on an explicit user request. Preview first; purge remains a dry run until --apply is present.

python3 /scripts/observe.py privacy purge
python3 /scripts/observe.py privacy purge --apply

Installing or invoking this skill must not edit AGENTS.md, CLAUDE.md, .claude/, .codex/, hooks, workflows, or any other file in the active repository.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Write a review

Versions

  • v0.1.0 Imported from the upstream source.

More in Agent Skills