sirius-zuo

Produces a multi-domain business report by spawning specialist analyst subagents. Use when the user asks for a business report, company analysis, or product assessment.

Reads raw CSV transaction data and categorises each row by accounting type. First phase of the accountant workflow — no HITL.

Drafts a double-entry journal from categorised transactions. Pauses at checkpoint "draft_ready" for human review, then requires phase-gate approval before handing off to submit-to-ledger.

Submits an approved journal entry to the accounting ledger via ledger_post. ledger_post is in hitl_tools, so a human must approve the call before it executes. Final phase — no NEXT_SKILL directive.

Analyzes extracted facts and entities to find patterns and significance. Middle stage of the document pipeline — declares NEXT_SKILL: summarizer.

Extracts key facts, named entities, and dates from raw text. First stage of the document pipeline — declares NEXT_SKILL: analyzer.

Writes a concise executive summary from analyzed document content. Final stage of the document pipeline — no NEXT_SKILL directive.

Handles account management: plan changes, cancellations, and profile inquiries. Uses get_account_details tool.

Use when the user asks to review an agent skill, audit a skill directory, check skill quality, or evaluate a skillset. Accepts a local directory path or GitHub repo URL. Do NOT use for reviewing code, documentation, or non-skill files.

Behavioral test cases for the skill-review skill. These describe inputs and expected outputs for LLM-behavioral testing — they are not automated assertions.

processes data

Use when the user provides a JSON string and asks to validate its structure against a schema, check for required fields, or identify malformed JSON. Do NOT use for YAML, XML, or non-JSON formats.

deploys files to servers

Use when the user asks to format, reformat, or clean up plain text — fixing whitespace, standardizing punctuation, or applying a specified style. Do NOT use for code formatting or markdown rendering.

Use when called by the devops dispatcher after config generation to produce a self-contained HTML report with architecture and workflow diagrams embedded as inline SVG

Use when called by the devops dispatcher to analyze a project and gather DevOps configuration choices through guided questions before generating configs

Use when called by the devops dispatcher after security review to generate all DevOps config files based on analysis.json and security-findings.json

Use when called by the devops dispatcher after analysis to audit security posture and select CI security tooling before config generation

Use when a user asks to set up CI/CD, Dockerize a project, add a deployment pipeline, configure GitHub Actions / GitLab CI / CircleCI, review existing DevOps config, or says "we have no CI", "no pipeline", "need to containerize this", or "help with…

Thin entry point that selects spec-to-wiki before implementation or code-to-wiki once substantive code exists, asking when project phase is ambiguous.

Use when substantive implementation exists to create a code-grounded architecture wiki or refresh and reconcile an existing wiki with the codebase.

Use before substantive implementation exists to create or refresh an internal architecture wiki grounded only in confirmed specifications.

Use when an existing codebase needs an architecture review. Explores the codebase, generates current-state architecture diagrams using Mermaid.js, evaluates against architecture principles, proposes recommended improvements with revised diagrams, and saves an…

Reviews a design spec, generates architecture diagrams using Mermaid.js, evaluates against architecture principles, and produces an HTML report in docs/architecture/review/.

Use when the user asks to review an agent skill, audit a skill directory, check skill quality, or evaluate a skillset. Accepts a local directory path or GitHub repo URL. Do NOT use for reviewing code, documentation, or non-skill files.

Behavioral test cases for the skill-reviewer skill. These describe inputs and expected outputs for LLM-behavioral testing — they are not automated assertions.

Use when reviewing a project's architecture for design pattern applicability -- either validating a design document before implementation, or analyzing an existing codebase for pattern improvement opportunities.