Capture what matters
Once authorized, capture bounded task summaries, explicit corrections and limited code snapshots, tied to the repository and revision.
Bounded captureDEEPSEEK HARNESS · COMMUNITY PLUGIN
Build traceable experience from test-verified Python fixes. Give DeepSeek Harness scoped, source-backed context for later, similar tasks.
dsh plugin --profile web add dsh-reforge@0.1.0-alpha.2
Independent checks qualify the repair source; applicability and verification records stay with the experience.
FROM REPAIR TO EXPERIENCE
Keep bounded evidence instead of a full conversation. Check what happened, then decide when the experience is applicable.
Once authorized, capture bounded task summaries, explicit corrections and limited code snapshots, tied to the repository and revision.
Bounded captureReplay original failures, independent acceptance and regressions in temporary workspaces. Turn completion and model self-reports are not acceptance.
Independent checksExtract a candidate with the selected model, then reverify it independently. Later requests receive scoped context only when conditions and relevance qualify.
Inspect, revise or deactivateONE SMALL FIX. A COMPLETE EXAMPLE.
Two minutes from a failing form to a verified fix and a stored experience. Then run the same example yourself.
The example includes the original project, explicit patch, external acceptance and regressions. Its passing results come from real test execution.
npm run example:customer-formClone the repository and install its dependencies first; run from the project root. See the example guide. No key or DSH is needed.
A CLEAR NEXT STEP
From a first run to data maintenance, follow instructions with concrete verification steps.
Try the example, configure read-only access, then decide whether to enable learning.
Use the exact package and host, the same profile, and explicit permissions.
Two exact hosts on Windows and Linux. Check the verified scope.
Connection failures, repository mismatches and an empty experience list.
YOUR PROJECT. YOUR DECISION.
Learning is off by default. The example and website do not need your model key. Understand execution, sharing and costs before enabling learning.
Backups, upgrades and recoveryA bounded synthetic real-model loop was verified. Both retrieval arms abstained in the later-task pilot, so no effect of using memory was measured. Verified source evidence does not establish future-task benefit. Read the results and limitations
FIX. VERIFY. REFORGE.
A complete example. Traceable sources. Permissions you control.