Menu
Per-channel review standards (GitHub, Discord, Twitter/X, blog, internal Slack). Apply after brand-enforcement and claims-risk pass.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Detect and prevent product-specific terminology contamination across skills/ and prompts/. Run after any bulk content update, skill import from another project, or agent config change. Applied as a shared skill across all agents.
When and how Architect delegates tasks to AO, Mechanic, and Designer.
Checklist and rules for reviewing deployment requests. Architect assesses but NEVER executes.
Rules for triaging new issues, classifying severity, and routing to the correct agent.
How to classify and respond to CI/pipeline failures.
Rules for detecting and managing stale issues and PR reviews.
| name | channel-standards |
| description | Per-channel review standards (GitHub, Discord, Twitter/X, blog, internal Slack). Apply after brand-enforcement and claims-risk pass. |
| metadata | {"version":"1.0.0","type":"always-active"} |
Different channels have different audiences, formats, and risk profiles. This skill defines what Reviewer checks beyond the universal brand + claims rules.
Audience: Developers evaluating the project, contributors, AI coding tools.
Checks:
| Check | Fail If |
|---|---|
| Technical accuracy | Code examples don't compile; commands don't work; paths don't resolve |
| Formatting consistency | Headings mix levels (## then #### without ###); code blocks missing language hint |
| Working links | Any broken link; any link to private/internal resources |
| Code examples runnable | Snippets use undefined variables, skip required imports, reference removed APIs |
| File path citations | Every packages/core/src/... path actually exists (grep the repo) |
| License header / attribution | All new files have AGPL-3.0 header if the org requires it |
| README sections in canonical order | Problem → Install → Usage → Architecture → Contributing → License |
Score penalty patterns:
npm tes instead of npm test) → -10Audience: Contributors, users, curious developers.
Checks:
| Check | Fail If |
|---|---|
| Tone | Casual-but-professional; no corporate marketing voice |
| Length | <2000 chars per message OR uses an embed (Discord embed limit applies) |
| Channel appropriateness | Technical depth matches channel (#support ≠ #engineering) |
| No internal info leak | No internal URLs, no internal agent names like "strategist", no credential references |
| Thread vs new channel | Long discussions use threads; one-liners use the main channel |
| No cross-post spam | Same content not repeated across >2 channels |
Discord-specific:
@here and @everyone blocked unless approved for a specific announcement typeAudience: Broad developer community, potential contributors, competitors watching.
Checks:
| Check | Fail If |
|---|---|
| Character limit | Over 280 per tweet; threads: each tweet self-contained |
| No thread-starting without approval | Unsolicited multi-tweet threads bypass this review; require Reviewer approval |
| Zero DeFi / token language | Any $, "yield", "stake", "bonding", etc. — block (see claims-risk) |
| Link present | Every informational post has at least one canonical link (github, yclaw.ai) |
| Handle consistency | @YClawAI or whatever the canonical handle is — check brand-voice.md |
| Hashtag policy | Max 2 hashtags per tweet; no #AI or other generic megatags |
X-specific risks:
Audience: Technical decision-makers, writers evaluating the project, archival readers.
Full editorial review required:
| Check | Fail If |
|---|---|
| Structure | Clear intro/body/conclusion; TOC for posts >2000 words |
| Accuracy | Every technical claim verifiable; code snippets tested |
| Voice | brand-enforcement voice attributes throughout, not just opener |
| Legal clearance | claims-risk passes for every paragraph (blog posts have more surface area than a tweet) |
| Sources | External claims have links; quoted data has source |
| Byline / attribution | Clear who wrote it — agent name or human name, not unattributed |
| SEO metadata | Title tag, meta description, OG image — all present and brand-voice-compliant |
Long-form posts require a human-in-the-loop before publish. Reviewer CAN approve the
voice/claims/legal pass, but final publish goes through a human maintainer — flag the
approval event with needs_human: true in payload.
Audience: The org itself — agents and Elon.
Lighter review. Main checks:
| Check | Fail If |
|---|---|
| Sensitive data leakage | Credentials, internal URLs, customer data in a channel that shouldn't have them |
| Cross-channel routing | Content about #yclaw-finance posted in #yclaw-development, etc. |
| Agent identity | Posted under correct agent's handle / avatar |
| PII | Any real user data — always flag |
Internal content does NOT need brand-voice or claims-risk screening. Those apply to external-facing content only. Reviewer's only job here is: "does this accidentally leak something that was meant to stay internal?"
| Content Type | Primary Channel | Reviewer Policy |
|---|---|---|
| Release announcement | GitHub release + X + Discord #announcements | Full review (GitHub + X + Discord checks) |
| Issue status update | GitHub issue comment | Internal-light review |
| Community question answer | Discord #support / #general | Discord review |
| Roadmap update | GitHub discussion OR blog | Full review + "not shipped" qualifier required |
| Bug report notice | GitHub issue + maybe Discord | Internal-light; redact credentials/IPs |
| Agent standup | Discord #yclaw-executive | Internal-light |
| External partnership comm | Private email / DM | Always escalate to Elon; Reviewer does not auto-approve |
After applying channel-specific checks:
brand-enforcement skill.claims-risk skill.oss-legal-guardrails skill.