- name
SKILL.md
Le fichier du skill reste dans le dépôt de l'éditeur.
Collez la commande copiée dans un terminal. Elle installe ce SKILL.md dans cet agent.
npx skills add omagents/omagents --skill hyperplan -a cursor -yCollez la commande copiée dans un terminal. Elle installe ce SKILL.md dans cet agent.
npx skills add omagents/omagents --skill hyperplan -a claude-code -yCollez la commande copiée dans un terminal. Elle installe ce SKILL.md dans cet agent.
npx skills add omagents/omagents --skill hyperplan -a codex -yCollez la commande copiée dans un terminal. Elle installe ce SKILL.md dans cet agent.
npx skills add omagents/omagents --skill hyperplan -a grok -yEnvoyez le message copié dans un chat Grok Bot. Le Bot récupère ce SKILL.md sur GitHub et l'ajoute aux skills que partagent tous vos Bots.
Install this skill: https://github.com/omagents/omagents/tree/HEAD/packages/core/skills/hyperplanDe l'éditeur
Dispatch 3 parallel critic subagents, each attacking a plan from a different perspective. The loop engine tracks completion; results are synthesized into a prioritized report.
loop_engine.py init hyperplan '[
{"critic": "security", "description": "Security critic: vulnerabilities, input validation, auth bypass, data exposure"},
{"critic": "architecture", "description": "Architecture critic: coupling, scalability, design patterns, maintainability"},
{"critic": "edge-cases", "description": "Edge cases critic: unhandled scenarios, error paths, race conditions, empty states"}
]'
Launch all 3 critics as background subagents in a single response:
task(background=true, description="Security critic", prompt="
You are a security critic. Review the following plan and find security issues.
Plan:
<insert plan here>
Analyze for:
1. Input validation gaps
2. Authentication/authorization bypass
3. Data exposure (logs, errors, responses)
4. Injection vulnerabilities
5. Dependency risks
Output: List of findings with severity (CRITICAL/HIGH/MEDIUM/LOW) and specific file/line references.
")
task(background=true, description="Architecture critic", prompt="
You are an architecture critic. Review the following plan and find design issues.
Plan:
<insert plan here>
Analyze for:
1. Tight coupling between modules
2. Scalability bottlenecks
3. Missing abstractions or over-engineering
4. Error handling strategy
5. Testability concerns
Output: List of findings with severity (CRITICAL/HIGH/MEDIUM/LOW) and specific design recommendations.
")
task(background=true, description="Edge cases critic", prompt="
You are an edge cases critic. Review the following plan and find unhandled scenarios.
Plan:
<insert plan here>
Analyze for:
1. Empty/null/undefined inputs
2. Concurrent access / race conditions
3. Network failures and timeouts
4. Resource exhaustion (memory, disk, connections)
5. Boundary conditions (0, max, negative, unicode)
Output: List of findings with severity (CRITICAL/HIGH/MEDIUM/LOW) and specific scenarios.
")
As each background task completes, record the result:
loop_engine.py complete hyperplan <id> "Found 3 issues: 1 CRITICAL, 2 MEDIUM"
If a critic fails to produce results:
loop_engine.py fail hyperplan <id> "Critic timed out"
When all critics are done:
loop_engine.py summary hyperplan
Generate a consolidated report:
## Hyperplan Review Report
### Summary
- Security: 1 CRITICAL, 2 MEDIUM
- Architecture: 0 CRITICAL, 3 MEDIUM
- Edge Cases: 0 CRITICAL, 1 HIGH, 2 LOW
### Critical Issues (must fix before implementation)
1. [SECURITY] SQL injection in search endpoint (src/api/search.ts:42)
Fix: Use parameterized queries
### High Priority (should fix before implementation)
2. [EDGE] No timeout on external API calls (src/services/external.ts:15)
Fix: Add 30s timeout with retry
### Medium Priority (address during implementation)
3. [ARCH] Search module directly imports from auth module
Fix: Introduce interface in shared/types.ts
4. [SECURITY] Error messages leak internal paths
Fix: Sanitize error responses
5. [ARCH] No retry strategy for failed jobs
Fix: Add exponential backoff in queue processor
6. [EDGE] Empty search query returns all records
Fix: Return empty result or validation error
### Low Priority (nice to have)
7. [EDGE] Unicode characters in usernames not handled
Fix: Add unicode normalization
8. [EDGE] Negative page numbers cause 500 error
Fix: Validate and return 400
Adversarial plan review with 3 parallel critics using loop engine tracking. Each critic attacks the plan from a different angle: security, architecture, and edge cases. Trigger when the user says 'hyp
---name: hyperplandescription: Adversarial plan review with 3 parallel critics using loop engine tracking. Each critic attacks the plan from…---# packages/core/skills/hyperplan/SKILL.mdVue d'ensemble
Adversarial plan review with 3 parallel critics using loop engine tracking. Each critic attacks the plan from a different angle: security, architecture, and edge cases. Trigger when the user says 'hyperplan', 'review plan', 'critique plan', or wants adversarial analysis of a design.
FAQ
Téléchargez packages/core/skills/hyperplan/SKILL.md auprès de l'éditeur, ou copiez le chemin et ouvrez le dépôt. PlugMCP répertorie le skill et n'héberge pas le fichier.
L'éditeur le maintient sur https://github.com/omagents/omagents.
Claude, Cursor, Codex et Grok Build peuvent lire un SKILL.md que vous ajoutez depuis le dépôt de l'éditeur. Grok Bot l'installe lorsque vous envoyez dans un chat le lien du dossier GitHub du skill.