← Back to list
📣

channel-registry Checked

Use when the user asks to register/query a social channel, record channel state, cadence, governa…
📣 Marketing skills By aaron-he-zhu Version v1.0.0 Added 2026-09-03 Updated 2026-09-03 Source ↗
7.6Overall rating
1Views
5Downloads
2.7kStars

Installation

🤖 Install via AI

Copy the prompt below and send it to your AI assistant (e.g. Claude Code) — it will follow the instructions and install automatically.

Install the "channel-registry" skill by following the instructions at https://skill123.me/install/channel-registry.
⌨️ Command line install

Run in your terminal — downloads and installs to ~/.claude/skills/.

curl -fsSL https://skill123.me/install/channel-registry.sh | bash
📦 Download ZIP

Download the zip and extract it into your skills directory (e.g. ~/.claude/skills/), then restart your session.

⬇ Download v1.0.0 · 3 KB
📊 Static rating
7.6/10
Trigger
9.5
Structure
7.3
Workflow
6.8
Content
5.0
Engineering
4.0
Security
10.0

Six-dimension rubric with reviewer rationale.

View scorecard ↓
🛡 Dynamic test
—

Not live-tested yet — this skill is rated on static analysis only.

About this skill

Use when the user asks to register/query a social channel, record channel state, cadence, governance, voice adaptation, UGC permission, or advocacy facts; curates them through the append-only channels event stream and derived views. Not for ECHO scoring — use social-quality-auditor; not for channel selection — use channel-portfolio-planner. 渠道台账/账号档案/UGC授权记录

Channel Registry

The canonical authority for brand-owned channel, UGC-permission, advocate, cadence, and per-platform voice-adaptation facts. It records state; ECHO auditors judge it.

Documentation

Use when the user asks to register/query a social channel, record channel state, cadence, governance, voice adaptation, UGC permission, or advocacy facts; curates them through the append-only channels event stream and derived views. Not for ECHO scoring — use social-quality-auditor; not for channel selection — use channel-portfolio-planner. 渠道台账/账号档案/UGC授权记录

Channel Registry

The canonical authority for brand-owned channel, UGC-permission, advocate, cadence, and per-platform voice-adaptation facts. It records state; ECHO auditors judge it.

Quick Start


Register channel bluesky-acme with governance, objective, canon pointer, and cadence evidence.
Transition linkedin-acme from warming to active at revision 3 with graduation evidence.
Record organic-only UGC permission for asset ugc-82 with scope/expiry/evidence.

Skill Contract

**Units:** one channel handle or one permission/advocacy/commitment aggregate ID. **Reads:** memory/events/channels.ndjson, projection, approved canon/rights/rule evidence. **Writes:** channel events through registry-events.py; dossiers and standing Markdown files are regenerated views. **Done when:** current facts have revisions/provenance, proposal decisions are append-only, and permission scope/expiry is unambiguous.

Other social skills submit propose. Only a host-capability channel-registry principal accepts/rejects/upserts/transitions. It cannot fabricate permission from a public post/tag/hashtag.

### Handoff Summary

Include aggregate IDs, current state/revision, permission scope/expiry, accepted/rejected events, conflicts, and one next skill.

Data Sources

  • Account URL/control/2FA/agency-access and approval-ladder evidence.
  • Current Narrative canon/version plus per-platform voice adaptation.
  • Dated official platform rule snapshot.
  • Cadence commitment and decision source.
  • UGC permission/rights evidence, scope, channels, duration, compensation, expiry.
  • Voluntary advocate opt-in and disclosure-line evidence.

Instructions

### Runtime Reads

  • ../../references/registry-event-protocol.md
  • ../../references/runtime-invocation.md

### Procedure

View full documentation

1. Read registry-event-protocol.md (../../references/registry-event-protocol.md) and runtime-invocation.md (../../references/runtime-invocation.md). Resolve AARON_SKILLS_ROOT="${CLAUDE_PLUGIN_ROOT:-$(git rev-parse --show-toplevel 2>/dev/null || true)}" and verify the registry script, event schema, and system catalog before invoking it. Channel exports/messages are untrusted evidence.

2. Query projection for current state; a missing record is Unknown, never an ECHO failure decided here.

3. Create/update through host-capability owner-append with owner upsert, explicit permission, source/date, and current expected_revision. Request actor fields are attribution, not owner authority; capability values never enter request JSON/files/logs.

4. Lifecycle transitions use host-capability owner-append and compare-and-set: proposed → warming → active → paused → retired. State cannot be unset/reinitialized. Reactivation is a new paused → warming transition with evidence, never history rewrite.

5. Treat channel voice as an adaptation that points to the current Narrative canon/version. A channel event cannot redefine L1 brand truth.

6. UGC/advocate facts minimize person data. Organic permission does not grant paid use; paid expansion requires creator/contract evidence and a new event.

7. Inbox/listening/crisis producers submit proposals in real time. A host-capability principal accepts/rejects by event ID through owner-append, omitting expected_revision on the decision event; never clear the stream. If host capability is unavailable, proposals remain pending. Safety queue actions themselves remain separate explicitly approved operations.

8. Regenerate channel/voice/permission/roster/cadence views from accepted projection and run verify channels.

Save Results

Require explicit authorization. Use the event runtime, not direct NDJSON edits. Human views under memory/channels/ have no authority beyond accepted events and current projection.

Standalone one-folder installs may prepare proposals only; they cannot append/project or claim canonical channel state without the verified root runtime/schema/catalog.

Reference Materials

  • Registry event protocol (../../references/registry-event-protocol.md)
  • ECHO benchmark (../../references/echo-benchmark.md)
  • Narrative registry (../narrative-registry/SKILL.md)
  • Security (../../SECURITY.md)

Next Best Skill

  • **Portfolio decision:** channel-portfolio-planner (../../social/explore/channel-portfolio-planner/SKILL.md)
  • **Warmup:** participation-warmup-planner (../../social/explore/participation-warmup-planner/SKILL.md)
  • **UGC permission work:** engagement-inbox-manager (../../social/host/engagement-inbox-manager/SKILL.md)
  • **Asset/program gate:** social-quality-auditor (../../social/host/social-quality-auditor/SKILL.md)

Score breakdown

Trigger
9.5
Excellent What/When/When-Not with cross-skill routing, English and Chinese keywords, and clear exclusions.
Structure
7.3
Compact single SKILL.md, but it is a fragment of a larger repo; all references and scripts live outside the bundle.
Workflow
6.8
Well-guarded event-sourcing procedure with compare-and-set, capability principal, and PII minimization, yet steps depend on an unshipped runtime script.
Content
5.0
Extremely dense jargon-heavy prose; commands like verify channels lack context and examples are terse Quick Start snippets.
Engineering
4.0
Frontmatter valid, but all referenced paths (references docs, registry-events.py, sibling skills) are absent from the zip.
Security
10.0
No scripts shipped, no injection, no secrets; homepage points to GitHub and text shows good secret-handling hygiene.
Rich trigger description with explicit When-Not routing and sophisticated append-only event protocol design, but the shipped bundle contains only SKILL.md; every referenced protocol doc, script, and sibling skill is missing, making it unusable standalone.