/install operational-framework
Operational Framework
A disciplined approach to implementing changes with full traceability.
When to Use
- Implementing improvements or new features
- Making configuration changes
- Any work that needs rollback capability
- Generating case study briefings
Architecture (generic)
\x3Cworkspace>/
├── decisions/ # Decision logs (JSON per day)
│ └── \x3Cdate>.json # Example: 2026‑05‑08.json
├── rollbacks/ # Snapshot directories
│ └── \x3Ctimestamp>/ # Example: 2026‑05‑08_2005/
├── briefings/ # Case‑study markdown files
│ └── \x3Cdate>.md # Example: 2026‑05‑08.md
└── TODO.md # Persistent task list
Replace \x3Cworkspace> with the root of your OpenClaw workspace (usually ~/.openclaw/workspace).
Decision Logging (generic)
Log each major decision before you start changing anything. Use a simple JSON schema:
{
"id": "dec_\x3Cdate>_\x3Cseq>",
"timestamp": "\x3CISO‑8601>",
"title": "\x3Cshort description>",
"context": "\x3Cwhy this decision matters>",
"options_considered": ["\x3Copt1>", "\x3Copt2>", "\x3Copt3>"],
"chosen": "\x3Cselected option>",
"reasoning": "\x3Crationale>",
"expected_outcome": "\x3Cwhat success looks like>",
"risk_mitigation": "\x3Chow to handle failure>",
"status": "pending|implemented|reverted"
}
How to log:
- Manually edit a file in
decisions/(e.g.,2026-05-08.json). - Or, if you have a CLI wrapper, run:
/decide "\x3Ctitle>" --context "\x3Cctx>" --options "opt1|opt2|opt3" --chosen "opt2" --reasoning "\x3Creason>"
Replace placeholders with your actual values.
Rollback System (generic)
Take a lightweight snapshot before you modify anything. The snapshot can be a simple copy of files or a git commit.
Typical workflow:
- Choose a name (e.g.,
2026-05-08_2005). - Copy the relevant files or the whole workspace into
rollbacks/\x3Cname>/. - Verify the copy.
- If needed, restore by copying back.
Example (shell‑style, adapt to your environment):
# Create snapshot directory
mkdir -p rollbacks/2026-05-08_2005
# Copy files you care about (or the whole workspace)
cp -r decisions rollbacks/2026-05-08_2005/
cp -r briefings rollbacks/2026-05-08_2005/
# ... add other paths as needed
Listing snapshots:
ls -1 rollbacks/
Restoring:
cp -r rollbacks/2026-05-08_2005/* \x3Cworkspace>/
The exact commands can be wrapped in a script for convenience.
Implementation Workflow
1. Decision Phase
- Log the decision with full context
- Define success criteria
- Identify rollback strategy
2. Snapshot Phase
- Capture current state
- Verify snapshot integrity
3. Implementation Phase
- Execute change
- Document as you go
- Test incrementally
4. Verification Phase
- Does it meet success criteria?
- Any unexpected side effects?
5. Briefing Phase
- Generate case study
- Note what worked/didn't
- Update TODO if follow-ups needed
Briefing Format
# Implementation Briefing: [Title]
**Date:** YYYY-MM-DD
**Decision ID:** dec_YYYY-MM-DD_XXX
## Context
[What triggered this]
## Decision
[What was decided and why]
## Implementation
[How it was implemented]
## Outcome
[Success/failure with evidence]
## Lessons Learned
- What worked well
- What would do differently
- Patterns to propagate
## Rollback Point
[Reference to snapshot if needed]
Quick Commands (examples)
Below are illustrative commands you can bind to your own CLI or script. They are not built‑in OpenClaw commands, but they show the typical flow.
| Action | Example Shell / Pseudo‑Command |
|---|---|
| Log decision | echo '{...}' >> decisions/$(date +%F).json |
| Create snapshot | ./snapshot.sh \x3Cname> (your wrapper script) |
| List snapshots | ls -1 rollbacks/ |
| Restore snapshot | ./restore.sh \x3Cname> |
| Generate briefing | ./brief.sh \x3Cdecision‑id> |
| Open TODO | vim TODO.md |
Feel free to adapt these to your preferred tooling (bash, Python, etc.).
TODO Integration
Maintain TODO.md in workspace root:
## 2026-05-08 Implementation Session
### Active
- [ ] Decision: Implement X
- [ ] Snapshot: AGENTS.md
### Completed
- [x] Decision: Add memory recall
- [x] Implemented: 2026-05-08
- [x] Briefing: briefings/2026-05-08.md
Key Principles
- Log before acting - Decisions documented before implementation
- Snapshot before change - Always have a rollback path
- Brief after completion - Document for future reference
- Never lose context - Everything survives session restarts
Integration
This framework integrates with:
- Self-Improving skill - Lessons feed into corrections.md
- HEARTBEAT.md - Periodic decision review
- AGENTS.md - Framework reference in operational procedures
Requirements
- No credentials required
- No extra binaries required
- Works with existing workspace structure
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install operational-framework - After installation, invoke the skill by name or use
/operational-framework - Provide required inputs per the skill's parameter spec and get structured output
What is Operational Framework?
A disciplined, reproducible workflow for AI agents to log decisions, create rollback snapshots, and generate briefings for any change or feature implementation. It is an AI Agent Skill for Claude Code / OpenClaw, with 59 downloads so far.
How do I install Operational Framework?
Run "/install operational-framework" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Operational Framework free?
Yes, Operational Framework is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Operational Framework support?
Operational Framework is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Operational Framework?
It is built and maintained by Tim (@tstokes06); the current version is v1.0.0.