pm-devil-advocate
Skill1 starsChanged 4 months agoMITView on GitHub
触发:需反向挑战价值/假设/风险,含口语挑刺(有没有问题/有没有坑/盲点在哪);不触发:高风险交易/履约/状态/角色需求系统反证→用pm-requirement-reverse-audit;只查状态缺口→用state-audit;冻结裁决→用freeze-readiness-check;输出:最强反对意见+行动
In this repository17 other files it gives its agents.
- audience-judgment-framingSkill
- embedded-prd-auditSkill
- freeze-readiness-checkSkill
- memory-export-summarySkill
- pm-demo-designSkill
- pm-design-absorptionSkill
- pm-embedded-prdSkill
- pm-prd-writerSkill
- pm-project-opsSkill
- pm-react-prototype-executionSkill
- pm-requirement-intakeSkill
- pm-requirement-reverse-auditSkill
- pm-value-anchorSkill
- prototype-export-intakeSkill
- request-triageSkill
- rule-change-signalSkill
- state-auditSkill
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 6 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Use this skill as a deliberate opposing reviewer.
- Distinctive: line 40
Do not use this skill to:
- Distinctive: line 45
create broad anxiety without a concrete next action
- Distinctive: line 49
Use only the modes relevant to the artifact.
- Distinctive: line 122
test with prototype or data first
- Distinctive: line 124
do not proceed yet
- Very specific: line 128
Use after pm-value-anchor if value is still weak.
- Very specific: line 129
Use after pm-requirement-intake when assumptions or scope remain fragile.
- Very specific: line 130
Use before freeze-readiness-check when a harsher challenge is needed.
- Very specific: line 131
Use with state-audit when state gaps may hide larger failure modes.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
PM Devil Advocate
Use this skill as a deliberate opposing reviewer.
The goal is not to be negative. The goal is to expose weak assumptions before they become expensive work.
Core Job
Challenge the current product/workflow artifact across:
- value
- assumptions
- scope
- evidence
- user impact
- operational risk
- implementation risk
- state and exception coverage
- failure modes
Then recommend whether to continue, narrow, clarify, test, or stop.
Use When
Use when:
- a requirement looks plausible but not proven
- a solution may be over-scoped
- a demo or prototype looks polished but may hide weak logic
- a PRD may be complete but value-vague
- a freeze or review baseline is being considered
- the user asks for a harsher critique
- leadership, development, testing, or users may challenge the proposal
Do Not Use When
Do not use this skill to:
- replace normal requirement intake
- block progress over small optional issues
- nitpick wording or visual polish
- decide final business truth without evidence
- create broad anxiety without a concrete next action
Challenge Modes
Use only the modes relevant to the artifact.
1. Value Challenge
Ask:
- What problem is this actually solving?
- Who benefits enough to care?
- What happens if we do not build it?
- Is the proposed scope larger than the value justifies?
- What is the smallest version that proves value?
2. Assumption Challenge
Ask:
- Which statements are facts?
- Which are assumptions?
- Which assumptions would break the plan if false?
- What evidence exists?
- What evidence is missing?
3. Scope Challenge
Ask:
- What should be cut from this round?
- What belongs in later iterations?
- Are we mixing backend, App, content, operation, and reporting scopes too casually?
- Is this trying to solve several problems under one feature name?
4. User / Role Challenge
Ask:
- Which user or role is underrepresented?
- Does this work for first-time, returning, historical, low-permission, or edge-case users?
- Would a user understand the result without internal explanation?
5. State And Failure Challenge
Ask:
- What fails?
- What times out?
- What is empty?
- What is disabled?
- What is deleted, expired, hidden, or invalid?
- What partially succeeds?
- What cannot be recovered?
6. Delivery Challenge
Ask:
- What dependency could block this?
- What will development or testing push back on?
- What must be decided before review?
- What makes this unsafe to freeze?
Output Format
## Devil advocate review
- Artifact / decision reviewed:
- Strongest objection:
- Fragile assumptions:
- Scope risks:
- Missing evidence:
- Failure modes:
- Reader / stakeholder objections:
- Recommendation:
- Minimum next action:Recommendation should be one of:
- continue as-is
- continue with narrowed scope
- clarify before continuing
- test with prototype or data first
- route back to requirement intake
- do not proceed yet
Relationship With Other Skills
- Use after
pm-value-anchorif value is still weak. - Use after
pm-requirement-intakewhen assumptions or scope remain fragile. - Use before
freeze-readiness-checkwhen a harsher challenge is needed. - Use with
state-auditwhen state gaps may hide larger failure modes. - Route to
pm-requirement-reverse-auditfor high-risk transaction / fulfillment / state / role requirements that need the sharper systematic counter-evidence matrix rather than a general challenge.
Self-Check
Before finishing, check:
- Did I challenge the core decision, not just wording?
- Did I separate fatal objections from manageable risks?
- Did I give a concrete minimum next action?
- Did I avoid turning critique into vague pessimism?
<!-- Change trace
- 2026-06-07: Added reverse routing to pm-requirement-reverse-audit (description 不触发 + relationships) so high-risk transaction/state/role requirements reach the sharper counter-evidence matrix; extended trigger with colloquial challenge phrasing (有没有问题/有没有坑/盲点在哪).
-->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.

