check-alignment
Skill8 starsChanged 14 days agoMITView on GitHub
Check if a proposed change aligns with mangrove-agent's stated principles, architecture, and trading-bot safety invariants. Use when the user describes a change and asks "does this fit", "check alignment", "review this approach", or "/check-alignment <description>". Read-only analysis - does NOT modify anything. Adapted from the Mangrove workspace `check-alignment` skill with trading-bot-specific documents.
In this repository8 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- audit-securitySkill
- connect-krakenSkill
- setup-krakenSkill
- setup-robinhoodSkill
- tool-specSkill
- trading-botSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 8 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
Check each principle if relevant:
- Distinctive: line 25
Every recommendation cites Mangrove intelligence
- Distinctive: line 36
Never expose plaintext keys or export capabilities to the agent.
- Distinctive: line 37
Default to testnet for new wallets.
- Distinctive: line 48
Never duplicate business logic across REST and MCP - service layer owns it.
- Distinctive: line 52
If the document exists, give a per-doc paragraph:
- Distinctive: line 61
If the document doesn't exist, note it and skip:
- Distinctive: line 72
Do NOT modify any file. Do NOT implement the proposed change.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Check Alignment
Parse $ARGUMENTS as a natural-language description of a proposed change - could be a new feature, a refactor, a dependency change, a workflow tweak.
Review the proposal against each of these documents. For every one that exists, state whether the proposal aligns or conflicts. If it conflicts, cite the specific principle or decision it violates and suggest a concrete adjustment.
Documents to check
1. SAGE.md - project identity + persona
- "What This Is" section: does the proposal still describe a local trading bot? Any scope drift toward cloud / hosted / multi-tenant breaks the v1 framing.
- "Architecture" section: dual-protocol (REST + MCP), three-tier access (free / auth / x402), service-layer pattern. Does the proposal respect these layers, or does it duplicate logic?
- "Project Context" + "Agent Identity": does the proposal change the user's project context or agent persona? Flag if yes - shared persona rules live in
SAGE.md; the user chooses the agent's name, saved locally throughset_agent_name. Identity changes are explicit, deliberate edits, not side effects of feature work.
2. .claude/rules/trading-bot-workflow.md - the 9 operating principles
Check each principle if relevant:
- Strategy-first, always (no manual-swap-by-default)
- Bulk candidate evaluation (no label-pick)
- Every recommendation cites Mangrove intelligence
- Paper before live
- Explicit confirmation at status transitions
- Small first allocation
- Wallet secrets NEVER in chat
- Signing validates intent inside isolated custody (
transaction_intentplus independent protected policy) - Testnet-first for learning (Base Sepolia 84532 by default)
Plus the stage machine (Stage 0 tour -> 1 Orient -> 2 Author -> 3 Review -> 4 Paper -> 4.5 Wallet -> 5 Live -> 6 Monitor). Does the change preserve the stage boundaries?
3. .claude/rules/wallet-presentation.md - wallet handling
- Never expose plaintext keys or export capabilities to the agent.
- Default to testnet for new wallets.
- Transaction signing routes through strict decoded intent validation and the separate-UID custody broker; x402, SIWX and marketplace proofs have their own narrow guards.
- Chain/router/function support and independently authorized wallet/spending policy may only expand through explicit review. Human import, backup and confirmation have no agent authority.
4. CONTRIBUTING.md - branching + PR flow
- Feature branch + PR for every change (no direct commits to main).
- Worktree workflows use
git -C <path>form so the hook recognizes the target. - CI must pass before merge.
5. docs/contributing.md - extension workflow
- Routes + services + MCP tools + tests mirror the repo structure.
- Never duplicate business logic across REST and MCP - service layer owns it.
For each document
If the document exists, give a per-doc paragraph:
### <doc name>
**Aligns** / **Conflicts** with <specific principle>
Reasoning: <one or two sentences>
Suggested adjustment (if conflict): <concrete change>If the document doesn't exist, note it and skip:
### <doc name>
Not present in this repo - skipping.Summary
End with a one-sentence verdict: Aligns, Conflicts, or Partially aligns (with which specific concerns).
Do NOT modify any file. Do NOT implement the proposed change. This is analysis only - the user decides whether to proceed after reading the alignment check.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

