agentsclimarketplace

Workbench hermes docs sync

Skill Fearvox/multica-ultimate-workbench/skills/workbench-hermes-docs-sync

Use when Hermes reviews Claude-authored writeups, public docs, skill changes, README/agent/issue-template changes, or upstream speed-match documentation before syncing or publishing.From its SKILL.md

Install
npx -y skills add Fearvox/multica-ultimate-workbench --skill workbench-hermes-docs-sync

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 23 stars23 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.9 KB, 841 tokens by cl100k_base, as published. Nobody here has run it

Workbench Hermes Docs Sync

Hermes is the second-pass reviewer for public-facing workbench documentation. Claude Code writes the first draft or patch. Hermes reviews it for coverage, consistency, public safety, and installability before sync or publish.

Core Rule

Claude writes. Hermes reviews. Supervisor accepts.

Hermes should not turn a docs-sync review into a rewrite unless the existing draft is materially wrong. Its job is to find every related public surface that must change, then return a bounded patch plan or PASS/FLAG/BLOCK verdict.

When To Use

Use this for:

  • skill additions, renames, removals, or registry-facing skill-map changes;
  • public README, docs map, agent roster, issue-template, autopilot, or install instruction changes;
  • Super.engineering/Hermes upstream speed-match docs;
  • any change that says "sync this everywhere", "public-facing", "skills.sh", "all Hermes", or "docs-sync";
  • release notes or PR text that must remain evidence-backed and public-safe.

Do not use this for private scratch notes, raw run logs, uncommitted screenshots, or live runtime mutation.

Public Surface Sweep

Hermes must check all potentially related public surfaces, not only the touched file:

  • README.md: install count, public overview, docs map, runtime model, Chinese overview when relevant.
  • AGENTS.md: read order, operating rules, file map, validation command list.
  • SYNTHESIS.md: current strategy when the change alters operating model.
  • DECISIONS.md: durable rationale for behavior-changing workflow changes.
  • skills/README.md: registry-facing skill list and install instructions.
  • skills/*/SKILL.md: affected skill contracts and cross-skill references.
  • docs/*.md: lane docs, public contracts, public-safe evidence boundaries.
  • agents/**/*.md: Hermes, Claude, Codex, remote, or reviewer role bindings.
  • issue-templates/ and autopilots/: reusable entry points affected by the new workflow.

If a surface is checked and not changed, name it in the review with "no change needed" and the reason.

Review Loop

  1. Read the Claude-authored diff first.
  2. Identify the behavior or public promise that changed.
  3. Sweep the public surfaces above using file search and direct reads.
  4. Flag stale install counts, dead links, missing skill-map rows, inconsistent role binding, or private/internal evidence leaking into public docs.
  5. Recommend the smallest patch set.
  6. Return a verdict. Do not silently mutate live Multica/Hermes runtime config.

Useful commands:

git diff --name-only
git diff --check
rg -n "skill-name|lane-name|Hermes|Claude Code|skills.sh|install" README.md AGENTS.md SYNTHESIS.md DECISIONS.md docs skills agents issue-templates autopilots

Verdict Contract

Return:

HERMES_DOCS_SYNC_REVIEW
source_author: Claude Code | other
changed_behavior:
public_surfaces_checked:
required_patches:
no_change_needed:
public_safety:
installability:
verification:
residual_risk:
VERDICT: PASS | FLAG | BLOCK

Use PASS only when every related public surface is either updated or explicitly judged unchanged. Use FLAG when publishable but missing optional proof. Use BLOCK when stale public instructions, broken install paths, private data exposure, or role-binding drift would mislead future users or agents.

Common Failure Modes

  • Only updating skills/README.md and forgetting README install count.
  • Adding a workflow doc without a decision record.
  • Adding a skill but not binding it to the Hermes role that should use it.
  • Treating Discord/terminal screenshots as public proof.
  • Updating AGENTS read order but forgetting the file map or validation list.
  • Letting Hermes rewrite the whole doc instead of reviewing the Claude draft.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,851. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.