verified-implementation
Skill1 starsChanged 11 days agoMITView on GitHub
Implement a planned change in small coherent increments, preserving repository conventions and verifying behavior as work progresses.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- adaptive-engineeringSkill
- code-reviewSkill
- existing-system-scanSkill
- memory-groundingSkill
- minimal-planningSkill
- simplification-gateSkill
- systematic-debuggingSkill
- task-triageSkill
- verificationSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
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 11
Avoid speculative extension points and generic abstractions.
- Distinctive: line 12
Avoid new dependencies unless existing/native options are materially worse and the trade-off is stated.
- Distinctive: line 13
Keep unrelated cleanup out of the diff unless it is necessary for correctness.
- Distinctive: line 14
Verify each meaningful boundary before expanding the change.
- Distinctive: line 15
If actual scope exceeds the complexity budget, stop and run the simplification gate.
- Distinctive: line 23
Prefer tests at the behavior boundary most likely to regress.
- Distinctive: line 26
test private implementation details merely to raise coverage
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Verified Implementation
Execution rules
- Make the smallest coherent change that satisfies the current acceptance criterion.
- Preserve existing architecture and conventions unless the task explicitly requires changing them.
- Avoid speculative extension points and generic abstractions.
- Avoid new dependencies unless existing/native options are materially worse and the trade-off is stated.
- Keep unrelated cleanup out of the diff unless it is necessary for correctness.
- Verify each meaningful boundary before expanding the change.
- If actual scope exceeds the complexity budget, stop and run the simplification gate.
Evidence-driven delivery
For STANDARD or DEEP behavior changes, follow the execution protocol: state a compact acceptance contract, reproduce failure or the intended behavior with a targeted test where useful, and verify the changed path. For DEEP/cross-module work, reconcile the planned impact map with the final diff. Delegate only independent, bounded tasks; hand over exact context and constraints, then inspect their diffs and results. Do not accept self-reported success as verification or authorize external side effects implicitly.
Testing
Prefer tests at the behavior boundary most likely to regress.
Do not:
- test private implementation details merely to raise coverage
- duplicate framework guarantees
- mock so aggressively that the real integration path is never exercised
For a bug, reproduce the failure before or alongside the fix whenever practical.
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.

