跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
其他技能

review-strategy

Diagnose what actual results and changed conditions mean for the current direction: check whether each bet ran with enough exposure, compare expected with observed, weigh competing explanations, recommend continue, increase, change, hold or close, and name the decisions to change at their owners. Use at a planned decision point, when a threshold is crossed, or when a consequential new observation arrives.

LeanOS-Technologies · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add LeanOS-Technologies/strategy --skill review-strategy

技能说明


name: review-strategy description: Diagnose what actual results and changed conditions mean for the current direction: check whether each bet ran with enough exposure, compare expected with observed, weigh competing explanations, recommend continue, increase, change, hold or close, and name the decisions to change at their owners. Use at a planned decision point, when a threshold is crossed, or when a consequential new observation arrives.

review-strategy

Writes strategy-review under the shared document rules.

When and what it decides

The question: what does this result actually tell us, about which choice, and what should change? A review diagnoses and recommends. The skills that own targets, offers, prices and focus revise their own documents.

Inputs and authority

  • venture-definition, objectives, bet-portfolio, focus-plan: what was intended and why.
  • Observations: readout, experiment-result, financial-report, customer or market observations, capability changes. Each carries its source. A plan, scenario or earlier proposal is never a result.
  • Execution facts: did the action happen, when, to whom, how often.

Method

Record sources, exact owner quotes and calculation premises in the relevant table’s basis or source cells. Carry decision status in its status column and the document’s role; prose states choices and ambition directly.

  1. Establish exposure first. For each bet: did the intervention run as designed, to the intended customers, for long enough to be read? If not, the disposition concerns execution, not strategy.
  2. Compare expected and observed. Use python3 workers/strategy/venture-strategist/tools/focus.py compare --expected "…" --observed "…" for quantities; it refuses incompatible units and unknown values. An unknown observation yields no conclusion, only a next observation.
  3. Generate competing explanations across the chain: target, offer, access, buying friction, delivery, execution, mechanism (objective trade-offs §5). For each material issue, name the observation that discriminates between explanations, and whether it exists.
  4. Locate the cause. Decide whether the result indicts one choice or a link of the venture's value chain itself (venture coherence §5). Only the latter calls for revising the venture-definition.
  5. Choose a disposition among continue, increase, change, hold, close, considering the remaining opportunity and the cost of the next commitment. Effort already spent is no reason to continue. A pivot is a change to core choices, justified only by a supported diagnosis.
  6. Name required changes by artifact and owning skill, with ordering and the next observation or decision condition. When evidence is insufficient, name the next observation and a bounded decision point instead of a verdict.

Evaluate and correct

DefectDetect byCorrect by
Verdict without exposureDisposition on a bet that barely ranChange to an execution finding; set the decision point
Single explanationDiagnosis lists one causeAdd the competing explanations and the discriminating observation
Plan counted as resultObserved column cites a projection or the bet's own expectationReplace with an observation or unknown
Review as authorityThe review states new prices, targets or focus itselfMove the change to Required changes for the owning skill
Sunk-cost continuationReason cites effort spentReason from remaining opportunity and next commitment

Run the group check; observation cells need basis.

Recovery

  • NeedData: the needed observation does not exist. Write the disposition hold, the observation needed, who produces it (Evidence) and the date of the decision point.
  • Conflict: two observations disagree. Name both sources and the check that would reconcile them; no disposition depends on either until reconciled.

Example

Illustrative fixture: fictional venture and observation records.

The paid-use result calls for completing the commercial intervention. It leaves the value and price questions available for a properly exposed test.

## Bet dispositions
| Bet | Objective | Expected | Observed | Execution and exposure | Disposition | Reason | Source |
| --- | --- | --- | --- | --- | --- | --- | --- |
| BET-1 | OBJ-2 | Paid use after the offer | No offer responses | Offer prepared but not sent | change | Complete delivery before interpreting buyer response | [observed: readout/2027-01#offer-delivery] |

## Diagnosis
| Observation | Explanations | Discriminating basis | Conclusion |
| --- | --- | --- | --- |
| Paid adoption has not begun | Offer did not reach users; price or value was rejected; payment route prevented purchase | [observed: readout/2027-01#offer-delivery] records that sending has not occurred | The evidence identifies an execution gap; buyer response requires delivery |

## Required changes
| Artifact | Owner | Change | Ordering | Next condition |
| --- | --- | --- | --- | --- |
| focus-plan | set-focus | Prepare ACT-2 for owner execution using an available payment route | After payment-route confirmation | Record actual delivery and responses |

The review supplies a correction to the focus owner and a concrete observation for the commercial choice.