Leitarkkk/dsh-research-nudge ↗★ 4
dsh-research-nudge
Research-debt guard for DeepSeek Harness: nudges coding agents to search docs, GitHub and the web before repeated local trial-and-error. 适合需要减少智能体本地盲目试错、提升研究深度的编码任务。
설치
$
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:Leitarkkk/dsh-research-nudgeConfiguration
The bundle inserts a row with the id research-nudge. Override that row in the profile's cordis.patch.yml:
- id: research-nudge
config:
enabled: true
debtThreshold: 20
maxToolCallsWithoutResearch: 15
maxMinutesWithoutResearch: 15
cooldownMinutes: 10
ordinaryToolDebt: 1
mutationDebt: 2
executionDebt: 1
failureDebt: 4
repeatedFailureDebt: 6
maxAgentSnoozeMinutes: 60
researchTools:
- web_search
- web_fetch
- github_search
- docs_search
- fetch_url
debug: false
DSH patch layers replace a row's entire config value rather than deep-merging it. Any omitted fields above fall back to this plugin's schema defaults. reminder may also be set to a custom string.
| Field | Default | Meaning |
|---|---|---|
enabled | true | Register the lifecycle listener |
debtThreshold | 20 | Debt score that makes a reminder eligible |
maxToolCallsWithoutResearch | 15 | Call-count fallback threshold |
maxMinutesWithoutResearch | 15 | Elapsed-time fallback threshold |
cooldownMinutes | 10 | Minimum time between reminders |
ordinaryToolDebt | 1 | Weight for other local tools |
mutationDebt | 2 | Weight for write/edit/delete-style tools |
executionDebt | 1 | Weight for shell/build/test-style tools |
failureDebt | 4 | Extra weight for a failed result |
repeatedFailureDebt | 6 | Extra weight for an equivalent consecutive failure |
maxAgentSnoozeMinutes | 60 | Upper bound for one agent-requested snooze of reminders |
researchTools | common web/docs/GitHub names | Substring patterns that reset state after normalization |
reminder | built-in advisory text | Model-visible reminder body |
debug | false | Log resets and queued reminders to stderr |