A collection of Claude Code plugins for automating git workflows, release documentation, and debugging.
Agent
Use this agent when the architecture-reviewer has returned a report with Critical or Major findings and the Mermaid diagrams need to be corrected before the browser preview is shown or a document is saved. Receives the review report, the diagrams.json path, and the requirements summary. Applies targeted fixes directly…
Agent
Use this agent only after implementation-planner has reported that an implementation plan was saved and confirmed. Reads the plan and the architecture document, then implements the project skeleton exactly as planned. Refuses to proceed if no valid, in-progress plan is provided.
Agent
Use this agent when the architecture-designer:design or architecture-designer:review skill needs to validate diagrams for technical correctness, cross-diagram consistency, requirements traceability, risk identification, operability/resilience, document/current-intent alignment, and (when applicable) Web3 or…
Agent
Use this agent when the architecture-designer:review skill's Step 2b (codebase-based review) needs to reconstruct the actual architecture of an existing codebase — for auditing a system against its (possibly stale or absent) architecture document, or for producing a first architecture document for a project that was…
Agent
Use this agent when the architecture-designer:design skill reaches stage 6 and needs database schema design, ERD creation, indexing plan, engine selection, transaction/concurrency-control strategy for high-contention entities, and secure connection configuration. Also used when a database-specific revision is needed…
Agent
Use this agent when the database-reviewer has returned Critical or Major findings and the database design needs targeted corrections before it is embedded in the architecture document. Receives the review report, the original database-designer output, the requirements summary, and the diagrams.json path. Applies the…
Agent
Use this agent after the architecture-designer:database-designer agent returns its output and before that output is embedded in the architecture document. Independently audits the database design for schema quality, normalization, ERD accuracy, index completeness, transaction/concurrency-control correctness for…
Agent
Use this agent when the document-reviewer has returned a DOCUMENT REVIEW FAILED verdict and specific format or content items need to be corrected before the document can be Approved. Receives the document path, the review report, and the requirements summary. Fixes the exact FAIL items in place and returns a fix log.…
Agent
Use this agent when the architecture-designer:design or architecture-designer:review skill has saved an architecture document and needs it audited for format compliance and content completeness before marking it Approved.
Agent
Use this agent when implementation-reviewer has returned an IMPLEMENTATION REVIEW FAILED verdict and specific code needs correcting before the implementation is presented to the user as finished. Receives the review report, the implementation plan path(s), and the architecture document path. Fixes exactly the flagged…
Agent
Use this agent when an approved architecture document exists and the user has confirmed they want to proceed with code implementation, before architecture-implementer is spawned. Resolves ambiguities, proposes a folder structure, waits for user confirmation, and saves the implementation plan file (or, for a large…
Agent
Use this agent after architecture-implementer (all parts, if the plan was split) reports Status Complete, to independently re-verify the generated code against the confirmed implementation plan and architecture document before the implementation is presented to the user as finished. Catches conformance gaps…
Agent
Use this agent when the architecture-designer:design or architecture-designer:review skill has opened the browser preview (Step 8 / step 4d) and wants to check whether diagrams actually render without visually overlapping elements — a real, rendered-geometry check using the chrome-devtools-mcp or firefox-devtools-mcp…
Agent
Use this agent when the user asks to "review changelog", "check changelog quality", "validate release notes", "is my changelog correct", or "review my CHANGELOG.md". Typical triggers include a user who has just generated a changelog and wants a quality check, a user who wants to verify the semver bump is correct…
Agent
Use this agent when the generate-release-notes skill encounters a character limit error ("Exceeds X char limit"), or when proactively validating that a planned command will fit within platform constraints before running it. Trigger when the script fails with a char limit error, the user asks to "validate release note…
Agent
Invoked by parallel-debug when two or more hypothesis-investigator agents both report a passing fix. Re-verifies cited evidence, checks for fix-diff file overlap, and returns ONEWINNER, MERGEFIXES, or ESCALATETOUSER. Do NOT invoke when a single hypothesis passes cleanly — apply that fix directly instead.
Agent
Spawned by the parallel-debug skill to investigate one specific root-cause hypothesis for a reported bug. Given a hypothesis, writes a failing test, investigates source code, applies a targeted fix, and iterates until the test passes or the iteration budget is exhausted. Returns a structured evidence report.
Agent
Use this agent when the user asks for a deep-dive analysis into a specific performance domain, says "deep dive into memory", "analyze this profiler output for me", "investigate the database bottleneck", "autonomous performance analysis", pastes large profiler data or GC logs for systematic single-domain investigation…
Agent
Use this agent when the user explicitly names the report-generator agent (e.g. "use the report-generator agent", "have report-generator produce this"), when another skill or agent needs a report generated as a delegated sub-step of a larger workflow without interactive back-and-forth, or when the user asks for a…
Agent
Use this agent when the /audit-fix command delegates post-fix verification, or when /audit-verify runs standalone review. Given the original audit report, a fix manifest from security-fixer, and the current file contents, independently verifies whether each fix actually closes the vulnerability at its root cause.…
Agent
Use this agent when the /audit command spawns it to perform read-only static analysis of a codebase. Typical triggers include being passed a project root path with a project map and asked to trace data flows from external inputs to dangerous sinks, being given entry point files and asked to verify authentication and…
Agent
Use this agent when the /audit-fix command delegates remediation work. Spawned only on explicit user request — never automatically. Given a list of finding-IDs from an existing audit report, fixes vulnerabilities one at a time: shows a fix plan first, waits for confirmation, then applies the minimal code change that…