メニュー
Generate a cross-platform AI audit (Trident) on a diff, brief, or artifact. Trigger: 'cross audit', 'Trident', 'second opinion', 'check with other AIs', 'cross-check this', 'get another perspective', /cross-audit
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Maintain canonical AGENTS.md (open spec). Trigger: 'agents.md', 'update AGENTS.md', or auto-fired by ijfw-team after agent generation.
First-class design intelligence. Dispatches to the best available design skill (ui-ux-pro-max, frontend-design, superpowers), then layers IJFW constraints on top. Triggers: 'design', 'redesign', 'UI', 'UX', 'dashboard', 'page', 'component', 'make it look better', 'polish', 'pretty', 'professional', 'user experience', 'layout', 'visual', 'mobile-first', 'dark mode', 'accessibility', 'colors', 'typography', 'brand'.
Check for and apply IJFW updates safely. Trigger: 'update ijfw', 'upgrade', 'latest version', 'is there a new version', /update
Use when the user says: 'build', 'create', 'plan', 'new project', 'brainstorm', 'design', 'UI', 'website', 'dashboard', 'app', 'help me build', 'launch', 'book', 'campaign', or anything project-level. Skill body decides Quick vs Deep path.
Check for and apply IJFW updates safely. Trigger: 'update ijfw', 'upgrade', 'latest version', 'is there a new version', /update
Compress memory/context files into terse form. Trigger: /compress, compress file
| name | ijfw-cross-audit |
| description | Generate a cross-platform AI audit (Trident) on a diff, brief, or artifact. Trigger: 'cross audit', 'Trident', 'second opinion', 'check with other AIs', 'cross-check this', 'get another perspective', /cross-audit |
Detect artifact. Accept: diff, file path, brief text, or HEAD~1..HEAD.
If none provided, ask once: What should I audit? (diff, file, or paste text)
Detect auditors. Check PATH for codex and gemini.
Default: one OpenAI-family + one Google-family, excluding caller's family.
Cap at 4 auditors total.
Dispatch in parallel via background bash -- never hand off to user. Prompt each: security findings, logic issues, reliability concerns, test gaps.
codex "Review for security, logic, reliability, test gaps: <artifact>" &
gemini "Review for security, logic, reliability, test gaps: <artifact>" &
wait
Reconcile. Deduplicate. Classify:
Emit report using the format below.
Any reconciliation report presenting multiple paths MUST use this structure:
VERDICT
<one-line converged recommendation>
OPTIONS
A -- <short-name>: <one-line what it is>
B -- <short-name>: <one-line what it is>
REVIEWER CONVERGENCE
<reviewer>: <letter> <score> "<their call>" => Option <letter> -- <name>
RECOMMENDATION
Option <X> -- <name> because <one-sentence why>.
NEXT ACTION
<exact command or step>
Never write "Option A" or "Option B" without "-- name" immediately after it on the same line.
Emit a gate-result block as the LAST content of your output. Nothing
after it. Use gate="cross-audit". Statuses: PASS | CONDITIONAL | WARN | FLAG | FAIL.
Format:
{
"schema_version": "1.0",
"gate": "cross-audit",
"status": "<STATUS>",
"project_type": "<from project-type-detector>",
"lenses": [],
"affected_artifacts": [],
"accounting": {"duration_ms": 0, "lenses_invoked": 0, "cost_usd": null},
"remediation": [],
"receipts_ref": null,
"supersedes": null,
"gate_id": "<gate-with-colons-replaced-by-dashes>-<ts>-<rand4>",
"emitted_at": "<ISO-8601>"
}