Agent Skill Diagnosis

Use only for exactly one Agent Skill target and one observed or measured anomaly. Direct selection is allowed. Automatic selection requires a target and incident evidence; generic terms such as

skill

debug

, or

performance

are insufficient.

This skill performs diagnosis only. It does not write the canonical skill, optimize the catalog, or own the final patch. Route a verified skill objective through

tk-learn

as the sole

create | improve | merge

writer. Do not semantically modify the canonical source skill.

Input Gate

Record:

exact target package/path, installed ref, origin, host, and invocation;
incident prompt, expected behavior or metric anchor, and observed result;
available transcript/event, file, Git, eval, or resource evidence;
known consumer override or host configuration.

Mark missing values as

unverified

. A missing incident or metric anchor does not mean

NotApplicable

. If fresh execution is unavailable or required evidence is inaccessible, the result is

Unverifiable | Blocked

, not permission to infer a cause.

Accept a

learn-ready

handoff only once when it specifies the incident, exact target, host/invocation, prompt, expected and observed result, and evidence. Do not repeat the same target + incident + blocker cycle.

Evidence

Before reproduction, check target provenance, description/body consistency, deterministic assertions, repository state, and adapter/host evidence. Then follow the workflow order for reproduction, the adjacent control, and any minimum experiment.

Classify reproduction as

Reproduced | Not reproduced | Inconclusive

. Self-report may suggest a hypothesis but does not prove root cause. Repeat a fresh run only when the first result is unstable or the boundary with the control is unclear. Do not require a fixed trial count, generic holdout suite, or rubric scoring when narrow evidence can determine the cause.

Read the following references only when applicable:

failure planes and evidence
empirical diagnostic method
upstream issue anonymization

Efficiency Gate

A resource claim requires a matched baseline, historical run, repository threshold, or explicit budget. Otherwise, record the observed value only as a profile and leave the direction

Unverifiable

. Lower token, time, call, retry, or fan-out usage does not offset a correctness or safety regression.

Workflow

Freeze: Fix the exact incident, target ref, must-preserve behavior, affected host, and reliable evidence/metric.
Reproduce: Reproduce once in a clean context and classify the result as
```
Reproduced | Not reproduced | Inconclusive
```
.
Control: Compare the nearest alternative. Distinguish loader from body, parent from child, candidate from grader, one host from another, and correctness from resource cost.

Isolate: Select one verified failure plane from:

selection | loading | instruction | planning | execution | formatting | evaluation | compatibility | efficiency | local override

Experiment only when needed: Run only in one run-owned isolated checkout. Change one root-cause theme and confirm or reject the cause. Do not treat the experiment as the canonical fix.
Route: Choose one owner based on verified evidence.

Stop at a conclusive cause or disposition. Allow a second experiment only if the first reveals a new concrete cause. Clean up only run-owned isolation.

🔴 CHECKPOINT / STOP · Next-Step Gate

Do not begin the next step, experiment, or handoff until each checkpoint passes.

Input checkpoint: An exact target, eligible Agent Skill incident, and incident evidence exist. If there is no eligible Agent Skill incident, such as an ordinary code bug, stop with
```
NotApplicable
```
. If a required incident or metric anchor is missing or unverified, stop with
```
Blocked | Unverifiable
```
, not
```
NotApplicable
```
.
Reproduction checkpoint: Record the fresh result as
```
Reproduced | Not reproduced | Inconclusive
```
. If
```
Inconclusive
```
, do not finalize the cause or route; stop with
```
Unverifiable
```
.
Isolation checkpoint: Evidence confirms one failure plane and an adjacent control that distinguishes it. Otherwise, make no root-cause claim and stop with
```
Unverifiable
```
.
Routing checkpoint: One concrete, testable objective and must-preserve boundary are verified. Otherwise, do not emit a
```
learn-ready
```
handoff.
🛑 STOP: After emitting
```
learn-ready
```
, wait for a separate explicit invocation of
```
tk-learn
```
.

Routing

Verified Skill Objective:

learn-ready

Use only when one existing package and one concrete, testable objective have been verified. Emit:

text

Target package: skills/<name>/
Objective: <one observable correction or cost reduction>
Evidence: <incident, control, code, event, or metric references>
Must preserve: <behavior, safety, routing, authority, and host boundaries>
Affected execution: <smallest fresh scenario that decides the objective>
Metric: <actual measurement, labeled proxy, or unavailable>
Incident: <stable ID or source reference>

이는 이후 명시적으로 실행하는

tk-learn

의 input이다.

Other Dispositions

```
learn-candidate
```
: A new independently useful skill is needed.
```
eval-owner
```
: The grader, fixture, harness, or assertion is the verified cause.
```
host-owner
```
: The loader, metadata, adapter, or host runtime is the verified cause.
```
local-only
```
: A consumer override/configuration causes the incident.
```
no-change
```
: Target behavior is correct or the incident is not reproduced.
```
unverifiable
```
: Evidence cannot safely determine the result.

In an external consumer repository, verify the upstream origin/ref and current upstream behavior before proposing an anonymized issue. Classify only a redacted proposal that has passed duplicate checking as

upstream-draft-ready

; do not automatically create, comment, label, or publish it.

Results

## Diagnosis

로 시작하고 이어서

## Action

을 출력한다. 필요할 때만

## Remaining uncertainty

를 추가한다.

하나의 incident에는 짧은 설명을 사용한다. 여러 symptom이 하나의 cause를 공유하면

ID | Incident | Root cause

형식으로 cause마다 안정적인

SD-##

row 하나를 유지한다. reproduction verdict, verified failure plane, evidence, route 및 정확한 next handoff를 보고한다. raw log, transcript, screenshot, secret 또는 반복된 run narration을 복사하지 않는다.

experiment evidence가 다섯 row를 넘거나 이후 resume에 정확한 reference가 필요하면

.tigerkit/skill-diagnosis.md

를 bounded incident IDs, candidate/control/holdout evidence refs, measurements 및 route와 함께 atomically replace한다. 채팅에는

## Diagnosis

## Action

및 필요한 uncertainty만 남긴다. archive, lifecycle state 또는 중복된 raw output은 만들지 않는다.

다음 terminal status 중 하나를 사용한다:

```
Pass
```
: diagnosis와 routing이 완료되었다.
```
Fail
```
: deterministic diagnosis/experiment claim이 gate를 위반했다.
```
Blocked
```
: 필요한 permission, decision 또는 environment를 사용할 수 없다.
```
Unverifiable
```
: provenance, reproduction, cause 또는 metric을 검증할 수 없다.
```
NotApplicable
```
: 적격한 Agent Skill incident가 없다.

Precautions

Do not assume the skill body is the cause.
Do not patch an unreproduced incident based only on wording intuition.
Do not change correctness, safety, or holdout behavior to reduce resource usage.
Do not use fixed repeated runs or judge majority to manufacture confidence.
Do not expose expected answers, secrets, or private evidence in prompts.

tk-skill-diagnose

NPX Install

Tags

SKILL.md Content