mcp-tool-extension
Skill3 starsChanged 9 days agoMITView on GitHub
Use when adding, changing, or removing an MCP tool in this repository. Covers catalog schemas, argument validation, executor dispatch, typed results, capability policy, query limits, tests, and documentation.
In this repository15 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- atomic-commitSkill
- mutation-lifecycleSkill
- odoo-compatSkill
- releaseSkill
- security-reviewSkill
- web-uiSkill
- atomic-commitSkill
- mcp-tool-extensionSkill
- mutation-lifecycleSkill
- odoo-compatSkill
- releaseSkill
- security-reviewSkill
- web-uiSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 6 are distinctive, and 8 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
Add the name to ToolName::ALL, as_str, TryFrom<&str>, and capability.
- Very specific: line 18
Add the JSON inputSchema in tool_definitions with descriptions, defaults, required fields, and bounded arrays/strings.
- Very specific: line 23
Add a typed argument struct in src/tools/arguments.rs following the existing naming and serde conventions.
- Very specific: line 29
Keep pagination limit/offset bounded by configured QueryLimits.
- Very specific: line 33
Add the dispatch arm in src/tools/executor.rs::execute_tool.
- Very specific: line 34
Use the OdooClient; never call Odoo directly from the MCP layer.
- Distinctive: line 46
Default-deny must remain intact.
- Distinctive: line 47
Never log, serialize, or return credentials or raw Odoo payload secrets.
- Distinctive: line 51
Add focused tests that do not require a live Odoo server:
- Very specific: line 61
Use src/test_support.rs and the existing mocked responses.
- Distinctive: line 65
Update the tool list and examples in README.md when behavior changes.
- Very specific: line 66
Update ROADMAP.md only when the item is defined there.
- Distinctive: line 67
Check version references only if the tool changes the advertised server capability or version.
- Distinctive: line 69
Run the full quality gate from AGENTS.md before committing.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
MCP Tool Extension
AGENTS.md is the canonical rule set. Follow this checklist for every tool
change.
1. Define the contract
- Choose a stable
odoo-*name and the matchingToolNamevariant insrc/tools/catalog.rs. - Add the name to
ToolName::ALL,as_str,TryFrom<&str>, andcapability. - Decide
requires_modelandrequires_client. - Add the JSON
inputSchemaintool_definitionswith descriptions, defaults, required fields, and bounded arrays/strings.
2. Validate before RPC
- Add a typed argument struct in
src/tools/arguments.rsfollowing the existing naming and serde conventions. - Reject missing, malformed, unknown, and oversized input before any Odoo call.
- Reuse
validate_domain_security,validate_field_names,validate_field_count,validate_order,resolve_limit, andvalidate_read_id_countwhere applicable. - Keep pagination
limit/offsetbounded by configuredQueryLimits.
3. Implement execution
- Add the dispatch arm in
src/tools/executor.rs::execute_tool. - Use the
OdooClient; never call Odoo directly from the MCP layer. - Return
ToolExecutionResultand map errors throughAppError; do not string-match Odoo errors. - For reads, use the pagination helpers and response-size protection.
- For mutations, do not add a second lifecycle path: route through
Operation/OperationExecutorand follow themutation-lifecycleskill.
4. Policy and safety
- Map the tool to the narrowest
Capabilityincatalog.rs. - Confirm instance, model, operation, field, and method policy scopes still apply.
- Default-deny must remain intact; never add an allow rule to make a test pass.
- Never log, serialize, or return credentials or raw Odoo payload secrets.
5. Tests
Add focused tests that do not require a live Odoo server:
- Schema: the tool definition exists, required fields are enforced, bounds are
correct, and the
ToolNameround-trip works. - Arguments: missing, malformed, unknown, and oversized input is rejected.
- Policy: allow and deny paths, including operation and model overrides.
- Executor: mock-server success, Odoo error mapping, pagination, and limits.
- Result: success and failure shapes; approval-required is not a transport error.
Use src/test_support.rs and the existing mocked responses.
6. Documentation and review
- Update the tool list and examples in
README.mdwhen behavior changes. - Update
ROADMAP.mdonly when the item is defined there. - Check version references only if the tool changes the advertised server capability or version.
- Run the full quality gate from
AGENTS.mdbefore committing.
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.

