Skip to content

feat(skills): Role 4 Risk Manager skill manifest#4

Merged
Brecht-H merged 1 commit into
mainfrom
orion-cc/role4-risk-manager-skill
May 17, 2026
Merged

feat(skills): Role 4 Risk Manager skill manifest#4
Brecht-H merged 1 commit into
mainfrom
orion-cc/role4-risk-manager-skill

Conversation

@Brecht-H
Copy link
Copy Markdown
Owner

@Brecht-H Brecht-H commented May 17, 2026

ADR-013A Role 4 — Risk Manager (Phase 1)

Kanban `t_role_riskmanager`. The `SKILL.md` manifest for the Risk Manager role —
the order-path veto enforcing the ADR-009 12-gate framework.

  • Model tier T3 — deterministic rules; the LLM is consulted only for a
    genuine novel market condition, and the deterministic verdict stands on any
    LLM error.
  • Paper-only, fail-closed (a gate failure denies the trade).
  • New skill dir `skills/trading/risk-manager/` — the home for the CORE-5 role skills.

The Python implementation (`risk.evaluate` / `risk.block`) ships in hermes-trader
(`src/hermes_trader/risk/risk_manager.py`, PR NousResearch#208) — per the ratified v0 contract
the skill module lives with the gate code it wraps; this manifest follows the
hermes-agent skill pattern.

Companion contract: `~/hermes-plan/HERMES_ROLE4_RISK_MANAGER_V0_CONTRACT_2026-05-17.md`

🤖 Generated with Claude Code

Summary by CodeRabbit

  • Documentation
    • Added comprehensive specification for the Risk Manager skill, covering deterministic order-path evaluation behavior, fail-closed requirements, input/output specifications, authority constraints, and practical usage examples for managing trading risk.

Review Change Stack

ADR-013A Role 4 (t_role_riskmanager), Phase 1. SKILL.md manifest for the
Risk Manager role — the order-path veto enforcing the ADR-009 12-gate
framework. Model tier T3 (deterministic rules; LLM only for novel
market-condition edge cases). Paper-only, fail-closed.

The Python implementation (risk.evaluate / risk.block) ships in
hermes-trader: src/hermes_trader/risk/risk_manager.py. Per the ratified
v0 contract, the skill module lives with the gate code it wraps; this
SKILL.md manifest follows the hermes-agent skill pattern.

Companion: ~/hermes-plan/HERMES_ROLE4_RISK_MANAGER_V0_CONTRACT_2026-05-17.md
Machine: orion-terminal

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@coderabbitai
Copy link
Copy Markdown

coderabbitai Bot commented May 17, 2026

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 818fb56e-d14f-4297-bf2d-990a2e2e0edb

📥 Commits

Reviewing files that changed from the base of the PR and between b8c14df and 60647ae.

📒 Files selected for processing (1)
  • skills/trading/risk-manager/SKILL.md

📝 Walkthrough

Walkthrough

This PR introduces a new Markdown specification for the Risk Manager skill, a deterministic 12-gate trading risk evaluation component that blocks, resizes, or allows orders within a paper-trading v0 environment. The document defines the skill's authority (veto-only), input contracts, execution interface, fail-closed safety rules, forbidden actions, and human escalation triggers.

Changes

Risk Manager Skill Specification

Layer / File(s) Summary
Skill metadata and authority
skills/trading/risk-manager/SKILL.md
Introduces the Risk Manager (Role 4) skill with version and platform metadata; defines its core 12-gate deterministic evaluation pipeline that produces ALLOW, RESIZE, or BLOCK verdicts, and establishes strict authority limits: veto-only (cannot approve orders), cannot override orchestrator global pause, paper-only in v0.
Input contract and prerequisites
skills/trading/risk-manager/SKILL.md
Specifies required prerequisites and all input payload elements needed for gate evaluation, including committee link fields, portfolio and account state (shadow orders, cooldown blocks, active restrictions).
Execution interface and output specification
skills/trading/risk-manager/SKILL.md
Documents execution via Python evaluate() and block() calls with code examples; defines GateVerdict output structure and field semantics for allowed vs. resize/block verdicts consumed by the order path.
Safety guarantees and forbidden actions
skills/trading/risk-manager/SKILL.md
Establishes mandatory fail-closed behavior for malformed inputs, schema violations, exceptions, and LLM/provider errors; enumerates forbidden actions (live submission, direct exchange/wallet access, schema mutations, unauthorized API spend, pause override, unblocking pipeline blocks).
Integration, escalation, and references
skills/trading/risk-manager/SKILL.md
Defines escalation triggers to human operators (drawdown/maxDD breaches, correlation violations, gate bypass attempts, simultaneous gate failures, novel market conditions); provides references to implementation files and v0 contract documentation.

🐰 A risk manager hops with care,
Twelve gates to check, all fair and square,
ALLOW or RESIZE, or BLOCK with might,
Fail-closed and safe—no trades in flight!
Then escalate when markets get tight. 🎯


🎯 2 (Simple) | ⏱️ ~12 minutes

✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests
  • Commit unit tests in branch orion-cc/role4-risk-manager-skill

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands and usage tips.

@Brecht-H Brecht-H marked this pull request as ready for review May 17, 2026 10:29
@Brecht-H Brecht-H merged commit 1e8ce3f into main May 17, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant