drift-report
Skill0 starsChanged 3 months agoMITView on GitHub
Use when checking whether design documents, READMEs, or specs still match the code they describe. Reads each document, locates the referenced source files, and reports only concrete divergences in a dated drift report. Detection only; does not fix code or docs.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- leiverkus/research-superpowerstheir own version, first seen Nov 1, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- leiverkus/research-superpowers★ 4same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository110 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/project-bootstrap/templates/agents.mdAGENTS.md
- a11y-auditSkill
- agentic-ragSkill
- agent-reachSkill
- ai3d-genSkill
- ai-coding-dictionarySkill
- ai-toolsSkill
- animation-directionSkill
- api-specSkill
- architecture-diagramsSkill
- architectureSkill
- ascii-artSkill
- ask-bundleSkill
- binary-triageSkill
- bragSkill
- brag-slimSkill
- career-opsSkill
- code-reviewSkill
- computer-useSkill
- context7Skill
- context-hygieneSkill
- creative-engineeringSkill
- cu-realtimeSkill
- data-analystSkill
- databaseSkill
- dead-wireSkill
- debuggingSkill
- deploySkill
- devin-configSkill
- dispatching-parallel-agentsSkill
- dmaSkill
- dockerSkill
- e2e-testingSkill
- edge-casesSkill
- executionSkill
- fact-checkSkill
- finishing-a-development-branchSkill
- firmware-probeSkill
- frontend-designSkill
- gatesSkill
- ghidraSkill
- ghSkill
- github-actionsSkill
- git-workflowsSkill
- grillingSkill
- handoffSkill
- humanizerSkill
- hyperframesSkill
- i18nSkill
- image-forensicsSkill
- impeccableSkill
- implement-layaSkill
- intakeSkill
- jiraSkill
- knowledge-modelingSkill
- laya-orchestratorSkill
- learn-modeSkill
- mcp-governanceSkill
- media-toolsSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 17
Default targets when the user names none: docs/ and .devin/plans/ in this bundle.
- Distinctive: line 25
If absent, ask the user to specify the target directory.
- Distinctive: line 29
Contextualize: read the document; extract intent, proposed changes, and specific filenames or symbols mentioned.
- Distinctive: line 35
Verify: read the current state of the source files with tools.
- Very specific: line 46
Report: write findings to <YYYY-MM-DD>_DRIFT_REPORT.md in the project .devin/ directory when it exists, else the design directory.
- Distinctive: line 61
Every claimed divergence cites the doc section and the code location actually read.
≈ “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
Drift Report (design vs code consistency)
Overview
Adapted from egm-drift-report in google/elephant-goldfish-model
(Apache-2.0). Detects where the codebase diverged from or failed to
implement what design documents describe. Report-only: divergences are
evidence for a human decision or a follow-up task, never silently
repaired in the same pass.
Default targets when the user names none: docs/ and .devin/plans/
in this bundle; in target projects, the directory of specs or READMEs
the user points at. spec-consistency.py covers plan-file internals
(checkboxes, declared refs); this skill covers doc-to-code fidelity.
Procedure
- Check input: require a directory of design documents or READMEs. If absent, ask the user to specify the target directory.
- List documents in the directory; process them sequentially to keep tool output under control.
- For EACH document:
- Contextualize: read the document; extract intent, proposed changes, and specific filenames or symbols mentioned.
- Locate source: find the referenced code via explicit paths, code search, or grep on the classes/functions/endpoints named. If the doc lacks links, tell the user to add them so future detection is cheaper.
- Verify: read the current state of the source files with tools; never compare from memory.
- Detect drift, looking specifically for:
- Missing features: logic described in the doc but absent in code.
- Incomplete models: dataclass/class/struct/proto fields the design specifies but the code lacks.
- Logic shortcuts: hardcoded values where dynamic logic was planned.
- Architectural divergence: implementation using different patterns or libraries than proposed.
- Filter: report ONLY differences. When code matches the design, record "No drift detected" for that document.
Report: write findings to
<YYYY-MM-DD>_DRIFT_REPORT.mdin the project.devin/directory when it exists, else the design directory. Strict format per document:## <DESIGN_DOC_NAME> <bulleted list of specific divergences> or "No drift detected"
Iron rules
- Report-only. Never edit code or docs to close a drift found in the same pass; record it and stop.
- Every claimed divergence cites the doc section and the code location actually read; no memory-based claims.
- "No drift detected" is a first-class result, not a failure.
- Be rigorous: a named object, field, or file absent from code is reportable drift; so is shipped logic unreferenced by any doc.
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.

