issueboss

Analyzes specific files identified by codebase-locator and explains HOW the code works, with precise file:line references. Traces data flow, identifies entry points, and surfaces implications and open questions relevant to the research topic. Use after codebase-locator has identified the relevant file paths.

Finds WHERE relevant code lives for a given feature, topic, or issue. Checks tokensave MCP first; falls back to Grep + Glob when unavailable. Returns a categorized file map (implementation, tests, config, docs, entry points) with no file reads. Use before dispatching codebase-analyzer to identify which files to examine.

Searches broadly for how things are *typically done* across the codebase — usage examples, established conventions, test patterns, and similar implementations. Runs independently; does NOT require codebase-locator output. Dispatched in parallel with codebase-analyzer by research-topic-processor.

Reads specific `.insights/` documents fully and extracts actionable intelligence: decisions, constraints, technical specs, gotchas, and open questions. Receives file paths from insights-locator or caller. Does NOT locate files. Processes at most 5 documents; notes additional ones in synthesis.

Finds relevant documents in `.insights/` for a given topic or issue slug. Scans `.insights/searchable/` using Grep and Glob, categorizes results by document type, and returns corrected paths (searchable/ stripped). Use before dispatching insights-analyzer to identify which docs to read.

Core per-topic research agent. Given a single research topic, classifies which research dimensions apply (code, pattern, insights, web), dispatches dimension agents in parallel, synthesizes their outputs inline, and returns { frontmatter, findings } using ---FRONTMATTER--- / ---FINDINGS--- / ---END--- delimiters. Does not know whether it is operating in issue-bound or ad-hoc mode.

IssueBoss research orchestrator agent. Dispatched by the issueboss skill (background by default). Receives an issue slug, iterates through uncovered ResearchTopic artifacts sequentially via the research-topic-processor agent, writes Research docs, attaches Research artifacts, then transitions to ResearchReview. Returns a brief summary distinguishing newly-researched vs already-covered topics.

IssueBoss triage agent. Dispatched by the issueboss skill (background by default). Receives an issue slug and optional context, performs full triage autonomously — reads config, investigates scope, writes triage doc, registers TriageResult artifact, transitions to TriageReview — then returns a brief summary to the caller.

Searches external web sources for a topic and returns synthesized findings with source URLs. Executes 2–3 searches before fetching; fetches 3–5 pages by default. If the caller prompt includes the tag `#web-deep`, raises the fetch limit to 8–10 pages. Use for current documentation, API references, ecosystem research, or any question requiring up-to-date information from the web.

Use when the user mentions issues, asks what to work on, references a backlog, asks about issue status or pipeline state, wants to find ready work, or references an issue slug (e.g. IB-42). Covers all IssueBoss MCP interactions including triage, artifact management, and pipeline advancement.

Ad-hoc research skill — investigate any topic (no IssueBoss issue required) by dispatching research-topic-processor, presenting findings inline, and optionally saving a document to .insights/shared/research/.

Orchestrate insights-locator and insights-analyzer to find and distill prior context from .insights/. Use when a topic may have prior research, at the start of brainstorming/spec/plan phases, or when the user says "check insights", "research insights", "look for prior context", or "what do we know about X".