Menu
Use when 3 consecutive core experiments or optimizations fail to meet expectations — forces honest status assessment and escalates the pivot/downgrade/kill decision to the user
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Use when designing the evaluation protocol for a method development project — locks primary metrics, datasets, seeds, statistical tests, and baseline list into an immutable contract
Use when G2 (plan freeze) and G3 (execution readiness) gates have passed — handles all implementation, baseline reproduction, experimentation, and iteration for both Type M and Type D projects; this is Phase 4
Use when research direction is validated (G1 passed) and the project needs method design, analysis framework, evaluation protocol, or story line — this is Phase 3 and must complete before any code or experiments
Reusable protocol for multi-agent discussions that iterate until convergence. Used by results-integration (Phase 5) and paper-writing (Phase 6) to ensure discussions actually resolve issues rather than just noting them.
Use when and ONLY when the user explicitly requests paper writing and G4 (write-ready) gate has passed — handles LaTeX structure, senior-level writing, reference verification, and iterative refinement
Use when Phase 1 (research-direction-exploration) is complete and G1 gate has passed — subjects the research problem to adversarial defense-style questioning, classifies project intent, re-evaluates venue target, and produces a final feasibility ruling before any method design begins
Baseado na classificação ocupacional SOC
| name | pivot-or-kill |
| description | Use when 3 consecutive core experiments or optimizations fail to meet expectations — forces honest status assessment and escalates the pivot/downgrade/kill decision to the user |
Repeated failure is signal, not noise. When 3 consecutive core experiments fail expectations, stop iterating and escalate. The agent does not get to decide whether to continue — the user does.
Core principle: Honesty over optimism. Escalate over iterate.
Violating the letter of this rule is violating the spirit of this rule.
AFTER 3 CONSECUTIVE FAILURES, STOP EXPERIMENTING. ESCALATE TO USER.
No fourth attempt. No "quick fix." Escalate.
digraph pivot_or_kill {
rankdir=TB;
detect [label="3 consecutive failures\ndetected" shape=doublecircle];
summarize [label="Summarize status\n(what was tried, what failed,\ntrend direction)" shape=box];
analyze [label="Analyze root cause" shape=box];
present [label="Present 3 options\n(Pivot / Downgrade / Kill)" shape=box];
decide [label="User chooses" shape=diamond];
pivot [label="Pivot\nChange method direction" shape=box style=filled fillcolor="#d4edda"];
downgrade [label="Downgrade\nLower venue target" shape=box style=filled fillcolor="#fff3cd"];
kill [label="Kill\nTerminate and archive" shape=box style=filled fillcolor="#f8d7da"];
detect -> summarize;
summarize -> analyze;
analyze -> present;
present -> decide;
decide -> pivot [label="pivot"];
decide -> downgrade [label="downgrade"];
decide -> kill [label="kill"];
}
State plainly:
Do NOT soften language. "Results declined across all three attempts" is correct. "Results showed some variability" is evasion.
Classify the failure into one of these categories:
| Category | Description | Signal |
|---|---|---|
| Method flaw | Core assumption is wrong | All variants fail similarly |
| Data issue | Insufficient, noisy, or wrong distribution | Results vary wildly or ceiling is low |
| Implementation bug | Code-level error | Should be caught by debugging, not this skill |
| Evaluation too strict | Unrealistic expectations | Baselines also fail to meet the bar |
Be specific. "Something is off" is not a root cause.
Present all three. No filtering, no pre-selecting.
a) Pivot — Change method direction.
b) Downgrade — Lower venue target.
c) Kill — Terminate the project.
THE AGENT CANNOT MAKE THIS DECISION. WAIT FOR USER.
Present the options. Stop. Do not nudge. Do not recommend. The user decides.
| Excuse | Reality |
|---|---|
| "One more try might work" | You said that twice already. Escalate. |
| "I think I see the issue now" | You thought that before too. Escalate. |
| "Killing the project is too drastic" | Killing saves resources for better projects. It's responsible, not drastic. |
| "Let me just tune hyperparameters" | Tuning is not innovation. If 3 approaches failed, the problem is deeper. |
| "The baseline is just very strong" | That's useful information. Maybe this gap isn't solvable with your approach. |
3 failures → stop → summarize → escalate → user decides
Honesty is not pessimism. Escalation is not failure. Continuing blindly is.