- Home
- Skills
- Agents & Orchestration
- Repository Investigation
Works with the AI Agents you already use
Repository Investigation
Conducts a read-only audit of a repository to verify ticket claims and document current architecture before planning.
$4.99
Repository Investigation
Example session with this skill installed
Investigate the repository for ticket PROJ-123: 'Add new error code for expired session in ErrorCode.java'. The ticket says it should go in the core error module.
- Read your context and instructions
- Compiled the repository investigation
I have completed the repository investigation for ticket PROJ-123. I located the relevant error handling components and verified the existing patterns for session-related exceptions.
repositoryContext
{
"inspectedFiles": [
{
"path": "backend/src/main/java/com/example/app/core/error/ErrorCode.java",
"relevantBecause": "Primary file mentioned in the ticket for adding new error codes."
},
{
"path": "backend/src/main/java/com/example/app/core/auth/SessionManager.java",
"relevantBecause": "Checked for existing session expiration logic to ensure consistency."
},
{
"path": "backend/src/test/java/com/example/app/core/error/ErrorCodeTest.java",
"relevantBecause": "Read to understand how new error codes are currently validated."
}
],
"existingConventions": [
"Error codes follow a prefixed naming convention: AUTH_*, DB_*, API_*.",
"ErrorCode enum entries include an HTTP status mapping and a default message."
],
"architectureBoundaries": [
"The core error module is shared across all backend services.",
"Changes here affect the global API response contract."
],
"verifiedFacts": [
"ErrorCode.java exists at the path specified and contains 45 existing codes.",
"There is currently no SESSION_EXPIRED or similar code in the enum.",
"SessionManager.java throws a generic AuthException when a session is null."
],
"assumptions": [
"The ticket implies the new code should follow the AUTH_ prefix, though not explicitly stated."
],
"notInspected": [
"Frontend error mapping components were not inspected as the ticket focuses on backend definitions."
],
"unresolvedQuestions": [
"Should we also update SessionManager to throw this specific new code, or just define the code for future use?",
"Does this new code require a corresponding localization update in the messages bundle?"
]
}
Next steps
- Confirm if the
AUTH_prefix is appropriate for the newSESSION_EXPIREDcode. - Decide if
SessionManagershould be refactored to use the new code immediately. - Proceed to implementation planning using the verified
ErrorCode.javastructure.
Connects securely to your tools. The creator never sees your data.
About this skill
Assumptions about a codebase lead to broken builds and wasted sprints. This skill forces a rigorous, read-only audit of the repository to establish what is true right now before a single line of code is written. It validates claims made in tickets against the actual files, tests, and architectural boundaries currently in place.
What it does
- Anchor extraction identifies package, class, and file names from ticket context to use as search targets.
- Deep discovery searches by behavior and feature area if literal names from the ticket are missing or renamed.
- Pattern identification reads neighboring implementations and tests to document existing conventions and abstractions.
- Boundary mapping notes whether changes impact core modules, feature-specific folders, or frontend/backend splits.
- Fact verification strictly separates verified findings from inferences to prevent implementation based on guesswork.
How it works
- Analyze input to extract concrete anchors from the ticket summary and description.
- Locate components by searching the local checkout for literal matches or behavioral equivalents.
- Audit surrounding code to understand established conventions and architecture boundaries.
- Generate report using the
repositoryContextschema, documenting facts, assumptions, and risks.
Frameworks & tools
This skill is language-agnostic and works with any repository structure. It relies on local file system access and basic search capabilities to inspect source code, tests, and configuration files.
Why this beats prompting it yourself
Manual prompting often results in the LLM hallucinating file contents or assuming a ticket's claims are accurate. This skill enforces a "read-only" constraint and a strict verifiedFacts vs assumptions classification that prevents planning on stale mental models.
Use cases
- Auditing a legacy codebase to find where a specific feature actually lives.
- Validating if a bug report's suggested fix location is the correct architectural boundary.
- Identifying reusable abstractions before implementing a new feature to avoid duplication.
Known limitations
Requires a locally checked out and readable repository. It cannot perform edits or create new files, as it is strictly restricted to read-only investigation.
Requires a locally checked out and readable repository. Does not perform edits, creations, or deletions. Limited to files that can be located via name or behavior search.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 18 days ago
Requires
Permissions
File Scopes
Read-only — never edits, creates, or deletes any file, regardless of what it finds.