Create a focused plan for keeping OpenClaw AMCS guidance in sync with the live AMCS MCP tool surface.
This should prevent skill drift as AMCS adds, renames, or restructures tools over time.
Problem
OpenClaw-facing skills or instructions for AMCS can go stale if they assume a static tool list. As AMCS grows, agents may:
miss new tools
keep referencing old tool names
route data into generic tools when a better structured tool now exists
lose confidence in AMCS integration if the documented surface drifts from reality
Goal
Add a discovery/sync mechanism so OpenClaw AMCS integration stays aware of the current tool list.
Proposed approaches
Option 1: Runtime discovery
Have OpenClaw inspect the live AMCS MCP tool list during startup or on demand, and use that as the source of truth.
Pros:
always current
no manual sync step
Cons:
needs a clean way to expose and consume tool discovery
may be too dynamic for some skill authoring flows
Option 2: Generated reference
Generate a compact reference file or skill resource from the live AMCS tool list.
Pros:
easy for skills to consume
reviewable and versionable
Cons:
still needs regeneration when tools change
Option 3: Drift check
Add a lightweight check that compares the documented AMCS tool guidance against the live AMCS MCP server tool list and reports mismatches.
Pros:
catches stale docs/skills
simpler than full runtime generation
Cons:
only detects drift; does not solve guidance updates automatically
Useful outputs
Any solution should ideally support:
current tool names
short descriptions
grouping by domain (thoughts, projects, files, skills, household, maintenance, calendar, meals, CRM, etc.)
identification of newly added tools
detection of removed or renamed tools
Acceptance criteria
There is a defined mechanism for discovering the live AMCS tool list
OpenClaw AMCS guidance can be checked or refreshed against that live list
Newly added AMCS tools can be surfaced without relying on memory alone
Drift between documented guidance and real tool surface can be detected
The plan is simple enough to maintain as AMCS evolves
Why this matters
This keeps the OpenClaw ↔ AMCS integration honest. If AMCS is evolving quickly, the agent guidance needs a way to keep up without turning into archaeology.
Closure note
Closing this because the core runtime discovery/sync problem is now largely addressed in the repo:
describe_tools exposes the live AMCS tool list
annotate_tool stores persistent usage notes tied to tools
llm/memory.md explicitly tells AMCS clients to call describe_tools at session start
There may still be value in a smaller OpenClaw-side authoring helper or drift-check workflow, but that belongs under the narrower OpenClaw integration issue rather than as a separate standalone ticket.
## Summary
Create a focused plan for keeping OpenClaw AMCS guidance in sync with the live AMCS MCP tool surface.
This should prevent skill drift as AMCS adds, renames, or restructures tools over time.
## Problem
OpenClaw-facing skills or instructions for AMCS can go stale if they assume a static tool list. As AMCS grows, agents may:
- miss new tools
- keep referencing old tool names
- route data into generic tools when a better structured tool now exists
- lose confidence in AMCS integration if the documented surface drifts from reality
## Goal
Add a discovery/sync mechanism so OpenClaw AMCS integration stays aware of the current tool list.
## Proposed approaches
### Option 1: Runtime discovery
Have OpenClaw inspect the live AMCS MCP tool list during startup or on demand, and use that as the source of truth.
Pros:
- always current
- no manual sync step
Cons:
- needs a clean way to expose and consume tool discovery
- may be too dynamic for some skill authoring flows
### Option 2: Generated reference
Generate a compact reference file or skill resource from the live AMCS tool list.
Pros:
- easy for skills to consume
- reviewable and versionable
Cons:
- still needs regeneration when tools change
### Option 3: Drift check
Add a lightweight check that compares the documented AMCS tool guidance against the live AMCS MCP server tool list and reports mismatches.
Pros:
- catches stale docs/skills
- simpler than full runtime generation
Cons:
- only detects drift; does not solve guidance updates automatically
## Useful outputs
Any solution should ideally support:
- current tool names
- short descriptions
- grouping by domain (thoughts, projects, files, skills, household, maintenance, calendar, meals, CRM, etc.)
- identification of newly added tools
- detection of removed or renamed tools
## Acceptance criteria
- [ ] There is a defined mechanism for discovering the live AMCS tool list
- [ ] OpenClaw AMCS guidance can be checked or refreshed against that live list
- [ ] Newly added AMCS tools can be surfaced without relying on memory alone
- [ ] Drift between documented guidance and real tool surface can be detected
- [ ] The plan is simple enough to maintain as AMCS evolves
## Why this matters
This keeps the OpenClaw ↔ AMCS integration honest. If AMCS is evolving quickly, the agent guidance needs a way to keep up without turning into archaeology.
---
## Closure note
Closing this because the core runtime discovery/sync problem is now largely addressed in the repo:
- `describe_tools` exposes the live AMCS tool list
- `annotate_tool` stores persistent usage notes tied to tools
- `llm/memory.md` explicitly tells AMCS clients to call `describe_tools` at session start
There may still be value in a smaller **OpenClaw-side authoring helper** or drift-check workflow, but that belongs under the narrower OpenClaw integration issue rather than as a separate standalone ticket.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Summary
Create a focused plan for keeping OpenClaw AMCS guidance in sync with the live AMCS MCP tool surface.
This should prevent skill drift as AMCS adds, renames, or restructures tools over time.
Problem
OpenClaw-facing skills or instructions for AMCS can go stale if they assume a static tool list. As AMCS grows, agents may:
Goal
Add a discovery/sync mechanism so OpenClaw AMCS integration stays aware of the current tool list.
Proposed approaches
Option 1: Runtime discovery
Have OpenClaw inspect the live AMCS MCP tool list during startup or on demand, and use that as the source of truth.
Pros:
Cons:
Option 2: Generated reference
Generate a compact reference file or skill resource from the live AMCS tool list.
Pros:
Cons:
Option 3: Drift check
Add a lightweight check that compares the documented AMCS tool guidance against the live AMCS MCP server tool list and reports mismatches.
Pros:
Cons:
Useful outputs
Any solution should ideally support:
Acceptance criteria
Why this matters
This keeps the OpenClaw ↔ AMCS integration honest. If AMCS is evolving quickly, the agent guidance needs a way to keep up without turning into archaeology.
Closure note
Closing this because the core runtime discovery/sync problem is now largely addressed in the repo:
describe_toolsexposes the live AMCS tool listannotate_toolstores persistent usage notes tied to toolsllm/memory.mdexplicitly tells AMCS clients to calldescribe_toolsat session startThere may still be value in a smaller OpenClaw-side authoring helper or drift-check workflow, but that belongs under the narrower OpenClaw integration issue rather than as a separate standalone ticket.