Stats

Actions

Tags

Help us improve

Share bugs, ideas, or general feedback.

consensus-engine | shannon

Skill

consensus-engine

Multi-validator agreement gate. N independent validators against same feature; synthesize confidence-scored verdict.

$

npx claudepluginhub krzemienski/shannon-framework --plugin shannon

Popularity

Stars

2

Invocation

How this skill is triggered — by the user, by Claude, or both

Slash command

/shannon:consensus-engine

User invocable

Model invocable

Inline context

Default effort

Context Preview

The summary Claude sees in its skill listing — used to decide when to auto-load this skill

Execution-time agreement gate. Spawns ≥2 (default 3) independent validators against the same feature; synthesizes confidence-scored verdict.

SKILL.md

69 lines · ~646 tokens

Similar Skills

ui-ux-pro-max

90.2k

Provides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.

context7-mcp

55.5k

Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.

context7-plugin

gitnexus-exploring

38.9k

Explores codebases via GitNexus: discover repos, query execution flows, trace processes, inspect symbol callers/callees, and review architecture.

1 file

Stats

LanguagePython

Stars2

MaintenanceExcellent

Last CommitMay 24, 2026

Actions

View Source View Plugin View on GitHub View README

Help us improve

Share bugs, ideas, or general feedback.

consensus-engine

Execution-time agreement gate. Spawns ≥2 (default 3) independent validators against the same feature; synthesizes confidence-scored verdict.

Behavior contract

Spawn N consensus-validator subagents in parallel via Task tool. Each owns an isolated evidence subdirectory:

e2e-evidence/consensus/<run-id>/
  validator-1/
  validator-2/
  validator-3/
  report.md  (synthesizer-owned)

Each validator runs the full journey list independently. Blind to peer verdicts.

Synthesize via consensus-synthesis (folded into this skill):

Per journey: compute (pass_count, fail_count, total).

Map to synthesis state:

State	Condition	Verdict	Confidence
UNANIMOUS_PASS	All PASS	PASS	HIGH
UNANIMOUS_FAIL	All FAIL	FAIL	HIGH
MAJORITY_PASS	≥⅔ PASS	PASS (after disagreement protocol)	MEDIUM
MAJORITY_FAIL	≥⅔ FAIL	FAIL (after disagreement protocol)	MEDIUM
SPLIT	Neither side ⅔	DISAGREEMENT_UNRESOLVED	LOW

Disagreement protocol (only on non-unanimous):
- Identify diverging criteria.
- Invoke sequential-analysis on the divergent evidence.
- Re-resolve via case (a) re-run missing validator, (b) re-run minority, (c) sharpen criterion (escalate to planner), (d) discard erroring validator.
- If unresolvable → SPLIT with LOW confidence; escalate.

File ownership (LOAD-BEARING)

Coordinator: writes NOTHING
Validator-N: writes ONLY to validator-<N>/ subdirectory
Synthesizer: writes ONLY report.md

A validator that writes outside its directory invalidates the run.

When to use

High-stakes change (security, payments, production data)
Pre-major-release validation
After a near-miss bug to harden the gate
User explicitly requests --mode consensus

When NOT to use

Routine refactor with no behavior change
Exploratory work
Speed-critical iteration (consensus is 3× the wall time of single-validator)

Iron rules

Confidence never upgrades — a late-arriving PASS verdict doesn't turn SPLIT into UNANIMOUS_PASS.
Evidence quality cannot substitute for agreement.
SPLIT is a real, reportable outcome. Do NOT silently downgrade to MAJORITY.