deep-investigation
Skill13 starsChanged 3 months agoMITView on GitHub
Investigate an operational problem to determine what happened, assess its impact, and recommend next steps using a hypothesis-backed approach.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- tttdddkkk/claude-skillstheir own version, first seen Oct 20, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- tttdddkkk/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository4 other files it gives its agents.
- AGENTS.mdAGENTS.md
- hypothesis-validationSkill
- on-call-handoffSkill
- suggest-follow-upsSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 8 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
Use available context first.
- Distinctive: line 28
After identifying what went wrong, generate 2-4 plausible root-cause hypotheses.
- Distinctive: line 47
Do not require a validated root cause.
- Distinctive: line 60
Do not infer customer impact from an error or alert alone.
- Distinctive: line 75
If the invoking prompt prescribes an exact output shape, follow it and fold the findings below into those sections rather than adding more …
- Distinctive: line 91
Do not assume the request represents a declared incident.
- Distinctive: line 92
Do not require a root cause before giving useful next steps.
- Distinctive: line 93
If the task is to validate one specific hypothesis, do not generate new hypotheses.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Evidence-Backed Investigation
Investigate an operational problem using relevant evidence to explain what happened, assess the impact, and identify useful next steps.
Procedure
1. Understand the request
Scan the conversation for:
- The reported symptom or unexpected behavior
- The affected service, environment, region, or workflow
- The relevant timestamps or time range
- Links to external context
- What the user needs to decide or do
Use available context first. Fetch additional context when it can change the conclusion or next action.
Perform a quick scan using available tools to determine if the problem is related to a recent, ongoing incident.
2. Generate hypotheses
After identifying what went wrong, generate 2-4 plausible root-cause hypotheses. Each must include:
- A short label
- Why the hypothesis is plausible based on the available context
- The signals that can validate or invalidate it
Prioritize the most plausible hypotheses and the signals most likely to distinguish between them.
3. Validate hypotheses and determine what happened
Investigate each hypothesis using the hypothesis-validation skill. For each hypothesis, determine whether it is validated, invalidated, or inconclusive and explain the conclusion with specific evidence.
Establish:
- What changed or failed
- When it started
- Whether it is ongoing, recovering, or resolved
- Which explanation is best supported by the evidence
Do not require a validated root cause. Clearly distinguish observed facts, supported explanations, and unknowns.
Record findings as you validate. Follow the Evidence rules in the agent instructions. Call evidence_record, evidence_remove, and evidence_list as those rules say.
4. Assess the impact
Determine:
- Which services, endpoints, regions, workflows, or users are affected
- How severe and widespread the effect is
- Whether the impact is increasing, stable, decreasing, or no longer present
- Which important impact could not be confirmed
Do not infer customer impact from an error or alert alone.
5. Choose next steps
Recommend 1-4 concrete actions based on the evidence. Prioritize actions that:
- Reduce current impact
- Confirm an important unknown
- Restore normal operation
- Prevent an unsafe or premature change
State what each action is expected to accomplish. Do not recommend broad follow-up work unless it helps with the current decision.
Result
If the invoking prompt prescribes an exact output shape, follow it and fold the findings below into those sections rather than adding more headings. Otherwise, answer with:
What happened
A short explanation of the observed behavior, timeline, current state, and best-supported cause. State important unknowns. Include related or recurring incidents only when this turn's evidence shows them. Do not state that no related incident exists because you do not have complete incident history to search.
Impact
The confirmed scope and severity. Clearly identify impact that could not be confirmed.
Next steps
A ranked list of 1-4 concrete actions for the responder.
Boundaries
- Do not assume the request represents a declared incident.
- Do not require a root cause before giving useful next steps.
- If the task is to validate one specific hypothesis, do not generate new hypotheses. Use the hypothesis-validation skill.
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.