| name | investigate |
| description | Investigate local repository, document, and attached-artifact context. Use for local investigation requests like "investigate", "find where", "understand this repo", "trace", and local-context research; do not use for web search. |
| license | MIT |
| tags | ["local-investigation","investigation","local-context"] |
| metadata | {"author":"Oleg Shulyakov","version":"1.1.0","source":"github.com/olegshulyakov/agent.md","catalog":"utility","category":"research"} |
investigate
Investigate local context and report evidence-backed findings.
Workflow
- Identify the target concept, behavior, file, command, error, or workflow.
- Search names, strings, docs, tests, configuration, and related symbols.
- Follow call paths, imports, references, generated sources, and tests only as needed.
- Distinguish live behavior from dead code, examples, fixtures, or stale docs.
- Summarize findings, gaps, and confidence with references.
Output
- Lead with the answer: state what was found or not found.
- Cite local evidence: include file paths, line references when available, and relevant commands.
- Separate inference: label deductions that are not directly stated in files.
- Name gaps: call out missing files, inaccessible artifacts, ambiguous ownership, or unverified runtime behavior.
- Keep scope tight: do not explain unrelated systems discovered during the search.
Boundaries
Scenario: Investigation stays local
Given the user asks for local investigation
Then search local files, project docs, attached artifacts, repository history, and available workspace context only
Scenario: Web research would be needed
Given the answer requires web search, browsing, or current-information research
Then do not perform it as part of this skill
And state that external research is outside the local investigation scope
Scenario: Findings are reported
Given findings are presented
Then ground them in file references, artifact references, command output, or clearly marked inference
Error Paths
Scenario: No matches are found
Given local searches return no matches
Then say what was searched
And suggest the next local search path
Scenario: Sources conflict
Given local sources disagree
Then prefer runtime wiring and tests over stale docs
And state the conflict
Scenario: Generated or external code is missing
Given a needed generated or external source is unavailable
Then identify the missing source
And explain how it affects confidence
Verification
Scenario: Output passes quality check
Given local findings have been produced
Then key searches were run before relying on memory
And implementation, tests, configs, or authoritative docs are preferred over secondary mentions
And no external or current-information claims are introduced
