Skip to content

zy839971925-zyy/Agent--deep-research-Workflow-Skills / integrations/deepseek-harness/skills/reasoning-workflow

reasoning-workflow

Verified

Default silent workflow for everyday questions, research, writing, design and engineering. Match effort to the request, resolve costly ambiguity, find specialist resources and verify the requested outcome. Explicitly selectable with “调用我的 workflow skills”.

FileKindSize
RELEASE.jsonconfig509 B
SKILL.mdskill md5.4 KB
UPGRADE.mdmarkdown13.1 KB
assets/icon.svgother3.1 KB
families/audit-verification.mdmarkdown1.1 KB
families/core-reasoning.mdmarkdown2.0 KB
families/decision-analysis.mdmarkdown698 B
families/deep-research.mdmarkdown2.2 KB
families/execution-control.mdmarkdown1.5 KB
families/workflow-learning.mdmarkdown500 B
references/change-governance-and-effectiveness.mdmarkdown12.0 KB
references/clarification-gate.mdmarkdown6.9 KB
references/comparative-synthesis.mdmarkdown3.2 KB
references/decision-and-recommendation.mdmarkdown6.6 KB
references/domain-patterns.mdmarkdown4.2 KB
references/engineering-delivery.mdmarkdown2.8 KB
references/evidence-and-provenance.mdmarkdown7.3 KB
references/formal-review-and-audit.mdmarkdown7.4 KB
references/harness-integration.mdmarkdown7.5 KB
references/harness-profiles.mdmarkdown11.8 KB
references/host-adapter.mdmarkdown11.8 KB
references/human-context-and-interpretation.mdmarkdown3.9 KB
references/hypotheses-and-bias-control.mdmarkdown6.2 KB
references/information-target-navigation.mdmarkdown7.4 KB
references/inquiry-and-research.mdmarkdown9.1 KB
references/measurement-and-operationalization.mdmarkdown3.1 KB
references/model-adaptation.mdmarkdown7.5 KB
references/multimodal-and-data.mdmarkdown3.4 KB
references/problem-framing-and-effort.mdmarkdown8.6 KB
references/reasoning-structure-and-decomposition.mdmarkdown5.5 KB
references/relationships-and-systems.mdmarkdown4.2 KB
references/research-modes-and-methods.mdmarkdown17.1 KB
references/research-to-report.mdmarkdown3.8 KB
references/retrieval-and-observation.mdmarkdown14.4 KB
references/runtime-and-delegation.mdmarkdown8.9 KB
references/semantic-state-contract.mdmarkdown5.7 KB
references/specialist-resources.mdmarkdown2.8 KB
references/state-model-and-invariants.mdmarkdown7.1 KB
references/synchronization-and-recovery.mdmarkdown9.6 KB
references/synthesis-execution-and-verification.mdmarkdown6.7 KB
references/task-strategies.mdmarkdown4.9 KB
references/time-scenarios-and-forecasting.mdmarkdown3.6 KB
references/traceability-and-integrity.mdmarkdown6.2 KB
references/ultra-research.mdmarkdown9.1 KB
references/worked-examples.mdmarkdown5.1 KB
references/workflow-learning-and-capa.mdmarkdown4.4 KB
references/writing-for-the-reader.mdmarkdown1.7 KB
routing-index.jsonconfig18.0 KB
schemas/capability-snapshot.schema.jsonconfig1.1 KB
schemas/checkpoint-manifest.schema.jsonconfig5.6 KB
schemas/delivery-manifest.schema.jsonconfig4.9 KB
schemas/dual-verification.schema.jsonconfig2.6 KB
schemas/edge-policy.jsonconfig3.8 KB
schemas/event.schema.jsonconfig1.9 KB
schemas/execution-node-state.schema.jsonconfig2.7 KB
schemas/execution-plan.schema.jsonconfig5.7 KB
schemas/execution-schedule.schema.jsonconfig3.7 KB
schemas/learning-record.schema.jsonconfig4.5 KB
schemas/raw-record.schema.jsonconfig3.3 KB
schemas/review-bundle.schema.jsonconfig1.6 KB
schemas/task-profile.schema.jsonconfig4.2 KB
schemas/work-state.schema.jsonconfig90.9 KB
schemas/worker-proposal.schema.jsonconfig3.1 KB
scripts/build_distribution.pyscript1.7 KB
scripts/check_report_evidence.pyscript3.6 KB
scripts/harness_bridge.pyscript9.2 KB
scripts/host_adapter.pyscript2.1 KB
scripts/learning_core.pyscript6.3 KB
scripts/prepare_task.pyscript2.6 KB
scripts/reference_core.pyscript1.4 KB
scripts/requirements.txtmanifest37 B
scripts/retrieve_learnings.pyscript1.6 KB
scripts/route_resources.pyscript4.1 KB
scripts/route_task.pyscript1.0 KB
scripts/routing_core.pyscript6.5 KB
scripts/runtime_core.pyscript36.4 KB
scripts/semantic_core.pyscript14.9 KB
scripts/validate_delivery.pyexecutablescript10.4 KB
scripts/validate_dual_verification.pyscript3.1 KB
scripts/validate_events.pyscript2.8 KB
scripts/validate_learning.pyscript718 B
scripts/validate_raw.pyscript3.0 KB
scripts/validate_runtime.pyexecutablescript2.5 KB
scripts/validate_state.pyexecutablescript3.3 KB
scripts/validate_workflow.pyscript9.0 KB

references/problem-framing-and-effort.md

View on GitHub
# Problem framing and effort

**Trigger:** Ambiguous, multi-objective, consequential, or poorly scoped tasks; any task where the wrong question would produce a misleading answer or action.

**Reads:** user request, supplied context, constraints, intended use, known success conditions.

**Updates:** epistemic task, objective, scope, answer conditions, pivotal questions, effort vector, lane selection.

**May invalidate:** premature research plans, proxy questions, over/under-scoped execution.

**Exit:** the actual question/outcome and proportionate effort are clear enough to proceed.

**Related:** [reasoning structure and decomposition](reasoning-structure-and-decomposition.md), [inquiry and research](inquiry-and-research.md), [human context and interpretation](human-context-and-interpretation.md), [state model and invariants](state-model-and-invariants.md), [clarification gate](clarification-gate.md).

**Return:** [root workflow](../SKILL.md) → Identify task / Calibrate effort.

Use this to decide what problem is actually being solved and how much cognitive machinery it deserves. Return the framing to the main workflow; do not turn this into visible ceremony unless it helps the user.

## Start from the epistemic task or outcome

“Answering a question” is a complete task, not a degraded form of project execution. First identify what kind of knowing or judgment the user is asking for: fact, explanation, causal mechanism, verification, comparison, diagnosis, forecast, interpretation, evaluation, recommendation, or direct reasoning. Only then decide whether creation/action is also part of the objective.

A prompt's nouns are not automatically the research structure. Identify the intended use: understand, interpret, persuade, create, choose, diagnose, forecast, verify, implement, or audit. Ask which unanswered questions can actually change that outcome.

For complex work, prefer a few **answer-bearing questions** over a generic topic checklist. Useful examples are questions whose different answers would reverse a recommendation, invalidate a design, change an explanation, or alter the evidence required.

A compact internal contract may include:

`objective | output/success | scope/time | constraints/authorization | pivotal questions | exclusions`

Infer routine missing detail. Before asking about a material gap, use the [clarification gate](clarification-gate.md) to check whether context, project evidence, external research, or safe derivation can resolve it. Ask only for the smallest user-owned input that can materially change the result; never infer consequential-action authority.


## Find the essence before forcing a category

When a task is unfamiliar or poorly framed, do not begin by assigning it to the nearest familiar taxonomy. Start from concrete observations/constraints, identify recurring patterns and consequential relationships, then infer the mechanism or essence that best explains why the problem behaves as it does. Use categories only when they improve compression, comparison, or routing after that understanding exists.

A useful check is: `If I removed the category labels, would I still understand what variables and mechanisms make this problem work?`

## Cognitive operators

Tasks can combine operators; do not force one label.

- **Interpret:** infer meaning from language, behavior, context, symbols, or artifacts.
- **Explain:** build a mechanism or coherent account of why/how.
- **Retrieve:** locate a fact, record, file, source, or observation.
- **Verify:** test whether a specific claim is supported and current.
- **Diagnose:** locate a fault or cause using discriminating observations.
- **Compare:** align alternatives on dimensions that matter to the user.
- **Decide:** recommend under goals, trade-offs, uncertainty, and reversibility.
- **Forecast:** reason from current state to conditional future states.
- **Create:** produce writing, code, design, media, plans, or artifacts.
- **Execute:** carry out an authorized action rather than merely describe it.
- **Audit:** attack an existing conclusion, artifact, methodology, or evidence chain.
- **Research/inquire:** discover the problem space, identify evidence needs, test alternatives, and reduce answer- or frame-uncertainty.

Operators compose. For example, “write an article” can be `create + explain + research` if factual dependencies matter, while “what did this sentence imply?” may be `interpret + compare explanations` with no external retrieval.

## Calibrate five kinds of effort

Treat effort as a vector rather than one scalar. A task can need deep reasoning but almost no change governance, or simple reasoning with strict governance because the action is high consequence.

### Reasoning breadth

Increase when multiple interpretations, stakeholders, mechanisms, jurisdictions, boundary conditions, or indirect effects can change the outcome.

### Evidence depth

Increase when facts are current, disputed, private, dispersed, difficult to observe, quantitative, or consequential. Reduce when the user's supplied material already determines the answer.

### Challenge depth

Increase for causal claims, accusations, personality/motive inference, forecasts, high-stakes choices, conflicting sources, and conclusions resting on a few assumptions.

### Verification depth

Increase when mistakes are costly or hard to reverse, when exact numbers/dates/versions matter, or when an artifact must actually work rather than merely look plausible.

### Governance depth

Increase when work changes persistent state, affects other people or systems, is difficult to reverse, crosses interfaces, requires authorization, or needs later monitoring. Governance depth controls traceability, change records, review independence, approval/readiness gates, rollout/rollback discipline, and effectiveness checks. Pure analysis or ephemeral drafts usually need little or none.

Use the host-selected model by default; do not infer model switching from a workflow depth request. Infer workflow depth from the task scenario rather than requiring a user label: simple/stable/directly answerable work is usually Light; a bounded lookup or modest synthesis is usually Standard; material multi-source, current, causal, or pre-build architecture work often merits Deep; Max/Ultra are for stronger frame instability, pivotal evidence gaps, broad uncertainty ecology, or explicit user steering. Treat these as revisable defaults, not a keyword classifier or compulsory profile. Raise or lower effort when evidence warrants.

User-specified depth controls all relevant workflow-effort dimensions and takes precedence over that inference. For example, “使用我的 workflow skills 的 ultra 深度去思考调研” explicitly selects Ultra for the research; preserve it. Ultra expands useful reasoning and research effort, not model identity, permissions, autonomy, persistence, output length, or quotas. Do not claim a model change. If the requested workflow resources or host capabilities are unavailable, use the strongest honest supported fallback and state any material limit; never silently relabel a shallower run as Ultra. “Maximum” means expand the useful epistemic frontier, not repeat searches or expose private reasoning.

## Escalation and de-escalation

Escalate when new evidence reveals hidden complexity, a pivotal contradiction, entity/version ambiguity, a strong causal leap, weak-signal/high-impact information, repeated search failure, or a consequential second-order relationship.

De-escalate when direct evidence settles the pivotal issue, remaining uncertainty cannot change the user's outcome, another route would only duplicate the same provenance, or execution now creates more value than additional analysis.

Do not equate long output with deep work. A deeply investigated task can still have a concise answer.

## Competing problem representations

When the framing itself is decision-sensitive, keep more than one plausible **problem representation** alive long enough to compare them. This is upstream of hypothesis competition: hypotheses ask what explains a problem inside one model; representation competition asks what kind of problem this is in the first place.

Do not force multiple frames for routine work. Use this only when different representations would change decomposition, evidence needs, causal interpretation, feasible actions, or the user's decision. Compare frames by what they explain, what they omit, what observations would distinguish them, and whether one frame merely restates a convenient proxy.

Prefer a provisional frame when evidence is incomplete. Collapse to one representation only after alternatives have low answer value or the differences no longer matter.

[← Return to root workflow](../SKILL.md)
brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).Stars: 322kUpdated Claude CodeCodexVerifiedInstall
brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).Stars: 198kUpdated Claude CodeCodexVerifiedInstall