WitnessCAUTION
🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
Overview
🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
ef7d4f0535e5OBSERVED · 2026-09-26Install
Commands as the repository documents them. They are shown, not run.
npm i @noble/ed25519
What it tells the agent
The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.
---
name: witness
description: Sign, verify, and track fix-marker regressions over time using a deterministic Ed25519 witness manifest. Works in any project — clone the toolkit, run init, register fixes, regen on each release.
argument-hint: "init|regen|verify|history [...]"
allowed-tools: Bash(node *), Read, Write, Edit
---
# Witness — cryptographic fix-regression tracking
The witness toolkit lets you ship every release with a *signed* manifest
that lists every documented fix in your codebase along with a sha256 +
marker substring. Anyone with the same git commit can re-derive the
public key and verify the signature without a committed private key.
A temporal history (JSONL) tracks how the fix population evolves across
releases — so when a regression appears, you can pinpoint *the commit
that introduced it*, not just "it's broken now."
This skill works two ways:
1. **Inside ruflo** — used by ruflo's own CI to gate publishes (see
`.github/workflows/v3-ci.yml` job `witness-verify`).
2. **In your own project** — copy `plugins/ruflo-core/scripts/witness/`
into your repo, run `init.mjs`, register your fixes in
`witness-fixes.json`, and call `regen.mjs` from your release pipeline.
## Quick start (any project)
```bash
# One-time bootstrap — creates verification.md.json,
# verification-history.jsonl, and witness-fixes.json template
node plugins/ruflo-core/scripts/witness/init.mjs --root .
# Edit witness-fixes.json: add { id, desc, file, marker } per fix.
# A "marker" is a distinctive substring that MUST appear in `file`
# while the fix is present. If someone reverts the fix, the marker
# disappears and `verify` reports it as `regressed`.
# Regenerate the manifest (signing requires @noble/ed25519)
npm i @noble/ed25519
node plugins/ruflo-core/scripts/witness/regen.mjs \
--manifest verification.md.json \
--history verification-history.jsonl \
--fixes witness-fixes.json
# Verify markers are present in the live tree
node plugins/ruflo-core/scripts/witness/verify.mjs \
--manifest verification.md.json
# Or authenticate the manifest and check source markers in a clean clone.
# Generated dist/ entries are explicitly reported as skipped.
node plugins/ruflo-core/scripts/witness/verify.mjs \
--manifest verification.md.json --source-only
```
## Temporal queries (ADR-103)
```bash
# Latest snapshot vs. previous
node plugins/ruflo-core/scripts/witness/history.mjs \
--history verification-history.jsonl summary
# For each currently-regressed fix, find the commit that introduced it
node plugins/ruflo-core/scripts/witness/history.mjs \
--history verification-history.jsonl regressions
# Status timeline for a specific fix
node plugins/ruflo-core/scripts/witness/history.mjs \
--history verification-history.jsonl timeline --id F1
# Machine-readable for CI
node plugins/ruflo-core/scripts/witness/history.mjs \
--history verification-history.jsonl summary --json
```
`summary` exits non-zero if any fix newly regressed since the last
snapshot — drop it in CI as a soft pre-merge gate.
## Anti-patterns
- **Hand-editing `verification.md.json`** — always regenerate via `regen.mjs`,
otherwise the signature breaks.
- **Markers that are too generic** (`'function'`, `'import'`) — pick something
unique enough that `grep` doesn't false-positive against unrelated code.
- **Skipping the history append** — without `--history`, you lose the
ability to bisect when a regression was introduced.
- **Committing one without the other** — `verification.md.json` and
`verification-history.jsonl` belong in the same commit; the JSONL is
what lets future you verify the signed manifest is the latest in the line.
## Files
- `scripts/witness/lib.mjs` — shared regenerate / history logic.
- `scripts/witness/regen.mjs` — CLI: sign + append history.
- `scripts/witness/history.mjs` — CLI: query the temporal log.
- `scripts/witness/init.mjs` — CLI: bootstrap into a fresh project.
- `scripts/witness/verify.mjs` — CLI: validate signature + markers.
## In ruflo's CI
`v3-ci.yml` job `witness-verify` runs after the behavioral smoke tests
and before `publish`. Failure modes:
| Failure | Cause |
|---|---|
| `signatureValid: no` | manifest hand-edited; re-run regen |
| `regressed: > 0` | a documented fix lost its marker since issuance |
| `missing: > 0` | a cited dist file no longer exists; rebuild or remove the entry |
| `scope: source-only` | signature + source markers checked; generated entries intentionally skipped |Trust audit
CAUTIONgrade B · trust 89/100 Install with care. The audit found things worth knowing before you trust its output.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | WARN |
| L1 | Static analysis of the code | NA |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (4)
crates
plugin/agents
plugin/commands
plugin/skills
Gates applied: no_behavioural_pass.
ef7d4f0535e5full audit observations/trust-audit/skill/ruvnet__witness.json · Report an issue / request a re-scanAudit history
Every audit this skill has had.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-09-26 | ef7d4f0535e5 | CAUTION | B | 89 | first audit |
Questions
What does the Witness skill do?
🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
Is Witness safe to install?
With care. The audit graded it B (89/100) and found 4 things worth knowing before you trust this skill, listed below with the exact line each was found on.
What can Witness access on my machine?
The audit observed no filesystem, network or shell use at all in its source.
How current is this page?
The grade is for one exact copy of the source (ef7d4f0535e5), read on 2026-09-26. The repository is watched, and a new audit runs when it changes — this is the first audit.