From tungnt-ai-skills
Use when investigating a bug, tracing an incident, exploring unfamiliar code, or building an evidence-backed case file before changing behavior
How this skill is triggered — by the user, by Claude, or both
Slash command
/tungnt-ai-skills:investigationThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Investigate before fixing. Reconstruct what is happening from evidence, document confidence, and stop at diagnosis unless the user explicitly asks for implementation.
Investigate before fixing. Reconstruct what is happening from evidence, document confidence, and stop at diagnosis unless the user explicitly asks for implementation.
Create or update a case file at docs/tungnt-ai-skills/investigations/<slug>.md for bugs, incidents, resumable investigations, or non-trivial code-area exploration. For a lightweight "explain this code path" request, report findings directly unless the user asks for a persistent case file.
The slug is a ticket ID when one exists. Otherwise derive a short lowercase kebab-case name from the problem statement. If the file already exists, resume it when the user asked for resume or follow-up; otherwise create <slug>-YYYY-MM-DD.md.
path:line, log timestamp, command output, or commit hash.path:line citations.Route the input.
Find the stronghold.
Initialize the case file.
Map the evidence perimeter.
Reason about cause.
Trace source where it matters.
Finalize.
quick-dev for trivial confirmed fixes, brainstorming plus writing-plans for ambiguous product or behavior choices, writing-plans only when requirements are already explicit, or requesting-code-review for review.# Investigation: <title>
## Hand-off Brief
1. **What happened.** <evidence-graded one-sentence problem statement>
2. **Where the case stands.** <status, strongest finding, remaining uncertainty>
3. **What's needed next.** <single recommended action>
## Case Info
| Field | Value |
| --- | --- |
| Ticket | <ticket ID or N/A> |
| Date opened | <YYYY-MM-DD> |
| Status | Active / Complete / Blocked / Superseded |
| Evidence sources | <logs, tests, commits, code paths, reports> |
## Problem Statement
<User-reported claim, refined or contradicted by evidence as needed.>
## Evidence Inventory
| Source | Status | Notes |
| --- | --- | --- |
| <source> | Available / Partial / Missing | <details> |
## Investigation Backlog
| # | Path to Explore | Priority | Status | Notes |
| --- | --- | --- | --- | --- |
| 1 | <description> | High / Medium / Low | Open | <context> |
## Timeline of Events
| Time | Event | Source | Confidence |
| --- | --- | --- | --- |
| <timestamp> | <event> | <citation> | Confirmed / Deduced |
## Confirmed Findings
### Finding 1: <title>
**Evidence:** <path:line, timestamp, command output, or commit hash>
**Detail:** <description>
## Deduced Conclusions
### Deduction 1: <title>
**Based on:** <confirmed findings>
**Reasoning:** <logical chain>
**Conclusion:** <what follows>
## Hypothesized Paths
### Hypothesis 1: <title>
**Status:** Open / Confirmed / Refuted
**Theory:** <description>
**Would confirm:** <specific evidence>
**Would refute:** <specific evidence>
**Resolution:** <what settled it, once known>
## Missing Evidence
| Gap | Impact | How to Obtain |
| --- | --- | --- |
| <gap> | <what it would resolve> | <collection step> |
## Source Code Trace
| Element | Detail |
| --- | --- |
| Error origin | <file:line or function> |
| Trigger | <what executes it> |
| Condition | <state that produces behavior> |
| Related files | <same code path> |
## Conclusion
**Confidence:** High / Medium / Low
<Summary separating Confirmed, Deduced, and Hypothesized conclusions.>
## Recommended Next Steps
### Fix direction
<Mechanism-level fix direction.>
### Diagnostic
<Additional confirmation steps if uncertainty remains.>
## Reproduction Plan
<Setup, trigger, expected result.>
## Side Findings
- <evidence-graded observation>
## Follow-up: <YYYY-MM-DD>
### New Evidence
### Additional Findings
### Updated Hypotheses
### Backlog Changes
### Updated Conclusion
npx claudepluginhub tungnt1405/tungnt-ai-skills-marketplace --plugin tungnt-ai-skillsGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.