Skip to content

Fleet Feature Catalog

Auto-generated by scripts/gen-features.mjs from the bundled skill/crew/role/workflow frontmatter. Do not edit by hand — run npm run build (or node scripts/gen-features.mjs) to regenerate.

A compact index of fleet capabilities shipped with this release. For full documentation see the other files in this docs/ directory.

Totals: 20 workflows · 29 skills · 58 roles · 7 crews (114 features)

Workflows

NameInvocationDescription
adversarial-research/research <topic or question>Adversarial deep research on a topic (parallel explorers + critic)
approval-gateDemo — build → human approval interrupt → deploy. Bundled example for #369 Phase 3C interrupts.
brainstormPre-feature brainstorming workflow. Spawns 3 parallel adversarial inquisitor agents (UX, technical, edge-cases) in round 1 to grill the user with questions until consensus is reached. Standalone (/brainstorm) writes 00-brainstorming.{md,context.md} in its own slug dir; chained via /feature --deep-brainstorm writes 01-brainstorming.md inline in the feature slug.
bug-autopilot/bug-autopilot <bug description or repro> [--scope diff|full] [--legacy-workflow-dispatch]End-to-end bug-fix autopilot: reproduce → fix → regression-test → code-review loop → open PR → wait for human approval → iterate on feedback → merge.
code-review/code-review [path or git-range]Run comprehensive multi-agent code review
code-review-autopilot/code-review-autopilot [target] [--scope=diff|full] [--ship] [--max-stage-visits <N>] [--legacy-workflow-dispatch]Iterative code-review auto-pilot: review → triage → fix → validate, looping until convergence; optional PR handoff via --ship. Pass --scope=full to audit the whole codebase instead of a diff.
compose-reviewCross-feature composition demo — fan-out per-file reviews → user approval interrupt → conditional apply/skip → report (#369 Phase 3A + 3B + 3C)
feature-pipeline7-phase feature planning pipeline (brainstorm → requirements → research → design → validation → DoD → tasks) — drives the /feature slash command
fixer-loop/fixer-loop [--max-iterations <N>] [--plan-dir <path>] [--legacy-workflow-dispatch]Auto-pilot loop that reviews, triages, fixes, and verifies. Iterates until convergence or budget exhausted.
freeformNo-op workflow for ad-hoc/freeform crew operation
full-review/full-review [git-range]Full-codebase report-only review (no fixes): audits the ENTIRE tree via the empty-tree..HEAD range with 5 parallel reviewers and writes CODE_REVIEW.md. Nothing is modified.
github-issue-worker-autopilot/github-issue-worker-autopilot [--max-stage-visits <N>] [--legacy-workflow-dispatch]GitHub-issues-driven autopilot: triage open issues, fix, validate with tests, generate proof artifacts, and ship as a draft PR.
init-investigation/init [refresh] [--legacy-workflow-dispatch]Workflow-driven /init — parallel investigators + critic + synthesizer
per-file-reviewDemo — fan-out per-file reviews and gather into a single digest (#369 Phase 3B)
planningMulti-step planning workflow with a bounded up-front grill, adversarial research, grill-derived DoD, task breakdown, and execution
security-autopilot/security-autopilot [target] [--ship] [--max-stage-visits <N>] [--legacy-workflow-dispatch]Defensive application-security audit autopilot: recon → hunt (per-class fan-out) → triage → PoC → fix → verify → report. Static analysis + local sandboxed PoCs only; never touches live/third-party systems. Pass --ship to open a PR with the fixes; default writes SUMMARY.md + report.md only.
security-binary-autopilot/security-binary-autopilot <target-binary-or-component> --subscription <isolated-sub-id> [--hours <n>] [--legacy-workflow-dispatch]Binary-analysis autopilot: target recon → static RE + patch-diff → coverage-guided fuzzing in ISOLATED-DISPOSABLE Azure VMs → crash triage + exploitability + minimal impact PoC → MSRC-format report → disclosure package. Authorized-scope only; isolated/disposable VMs; coordinated disclosure via MSRC only; no weaponization.
security-disclosure-autopilot/security-disclosure-autopilot [target] --subscription <isolated-sub-id> [--scope diff|full] [--fix] [--no-teardown] [--legacy-workflow-dispatch]Coordinated-disclosure autopilot: recon → scan → triage → local PoC → ISOLATED-DISPOSABLE Azure-VM detonation + evidence capture → MSRC-format report → optional fix → disclosure package. Authorized-scope only; MSRC coordinated disclosure only; isolated/disposable VMs (no prod/third-party, auto-teardown); evidence integrity (hash+timestamp).
subgraph-exampleDemo — parent workflow that invokes code-review as a subgraph stage
triageInvestigate then branch — fix / review / skip — based on artifacts, placeholders, and ifResult

Skills

NameInvocationDescription
agents-fleet-conventionsShared conventions for agents-fleet codebase (TypeScript/ESM/branded types/named exports)
ci-pipeline-authoringAuthoring robust CI/CD pipelines — test, coverage, regression matrix, and build gates (GitHub Actions)
coordinator-worker-architectureagents-fleet coordinator-worker primitives, boundaries, and concurrency
html-report-authoringAuthor polished, self-contained, interactive HTML reports in the agents-fleet light brand theme (SVG/PNG/interactive diagrams, never ASCII) with on-demand PDF export
hub-tool-disciplineHub topology — workers MUST route ALL coordinator-bound communication through report_to_coordinator
ponytail-minimalismYAGNI ladder and minimalism rules — forces the laziest solution that actually works
ponytail-reviewOver-engineering code review — finds what to delete, replace with stdlib/native, or shrink
prepared-statements-patternIntelDatabase prepared-statement conventions for better-sqlite3 (WAL, sync, idempotent migrations)
react-ink-patternsReact 18 + Ink 5 terminal-UI patterns — function components, hooks, Static, Box layout, polling
sec-authz-auditAudit authentication and authorization flaws — broken access control, IDOR, missing function-level authz, JWT weaknesses, session/cookie flaws, privilege escalation
sec-azure-vm-provisioningSafely provision an ISOLATED, DISPOSABLE Azure VM for PoC validation with the az CLI — dedicated resource group, isolated vnet/subnet, deny-all NSG, no public IP, snapshot before detonation, TTL tag, and guaranteed teardown; NEVER a production subscription
sec-binary-static-analysisStatic reverse engineering of a target binary (PE/ELF/Mach-O) — triage the file, extract imports/exports/strings/sections, disassemble + decompile (Ghidra headless / IDA / radare2 / objdump), identify dangerous APIs, parsing routines, and trust boundaries, and patch-diff a patched-vs-unpatched build to locate the fixed/target code
sec-cicd-auditAudit CI/CD and GitHub Actions security — over-privileged tokens, untrusted pull_request_target, script injection via github.event, unpinned actions, secret exposure, artifact tampering
sec-crash-triage-exploitabilityTriage fuzzing crashes and assess realistic exploitability — reproduce, classify (WinDbg !exploitable / !analyze, ASan reports), determine the memory-safety primitive (OOB read/write, UAF, type confusion), assess mitigations (DEP/ASLR/CFG/stack cookies), and write the minimal PoC that PROVES impact (controlled crash/AV) — never a weaponized exploit
sec-crypto-auditAudit cryptographic misuse — weak algorithms, hardcoded keys/IVs, insecure randomness, unauthenticated encryption, improper TLS/cert validation
sec-deserialization-auditAudit insecure deserialization, JS prototype pollution, and unsafe reflection across Node/TS and C#/.NET
sec-finding-reportDocument a security finding — title, qualitative severity, affected file:line, description, PoC/repro, impact, remediation, and framework mapping (CWE, OWASP Top 10 2021, MITRE ATT&CK)
sec-fuzzingCoverage-guided fuzzing of a target (parser/decoder/IPC/file-format) — build a harness (libFuzzer/AFL++/WinAFL/honggfuzz), assemble a seed corpus + dictionaries, run with sanitizers (ASan/UBSan), measure coverage, and minimize crashers — all inside an isolated, disposable VM
sec-injection-auditFind and safely confirm injection classes — SQL/NoSQL, OS command, SSTI, path traversal, SSRF, XXE, open redirect, header injection
sec-llm-agent-auditAudit LLM and agent security (MITRE ATLAS) — prompt injection (direct/indirect), tool/command injection via model output, system-prompt leakage, RAG poisoning, unsafe MCP/tool allowlists, over-broad agent permissions
sec-msrc-reportProduce an MSRC-format coordinated-disclosure report — affected product/build, vulnerability type + CWE, CVSS 3.1 and 4.0 vectors + base scores, attack vector/prerequisites/privileges/user-interaction, numbered repro steps, PoC reference, impact classification, affected configurations, suggested mitigation, and coordinated-disclosure timeline
sec-poc-authoringProve a security finding SAFELY — write a local failing test or minimal sandboxed proof-of-concept against the operator's own code only, that goes green once fixed
sec-recon-attack-surfaceMethodology to map a codebase's attack surface — entry points, trust boundaries, dangerous sinks, dependencies, CI/CD, and secrets locations
sec-secrets-auditDetect secrets and credential exposure in source, config, and git history — API keys, tokens, private keys, connection strings — with gitleaks-style patterns and remediation
sec-supplychain-auditAudit dependency and supply-chain risk — known-vulnerable packages, lockfile integrity, dependency confusion, typosquatting, SBOM review for npm and NuGet
sec-vm-detonationDetonate a PoC inside an isolated, disposable Azure VM and capture evidence — transfer via az vm run-command / Custom Script Extension, execute, collect logs/screenshots/crash-dumps/process+event evidence, hash + timestamp each artifact, assert the security impact vs a benign baseline, then snapshot-revert or delete the RG
security-audit-checklistThreat-model audit checklist — path traversal, SQL injection, prototype pollution, credentials, command injection, MCP allowlists
vitest-patternsVitest 4.x test patterns — fixtures, mocking, async, tmpDir cleanup
workflow-authoringWorkflow authoring guide — file anatomy, runner types, edges, forEach, postChecks, dispatch ceremony, author→validate→fix loop

Roles

NameTypeDescription
architecture-revieweragent_type: reviewerArchitecture and design pattern reviewer
context-researcheragent_type: explorerContext and dependencies researcher - investigates surrounding context and historical background
correctness-revieweragent_type: reviewerLogic and correctness focused code reviewer
counter-researcheragent_type: explorerCounter-arguments and alternatives researcher - investigates downsides and alternative approaches
default-coderagent_type: coderDefault system prompt for coder worker agents
default-coordinatoragent_type: coordinatorDefault coordinator system prompt for fleet orchestration
default-exploreragent_type: explorerDefault system prompt for explorer worker agents
default-general-purposeagent_type: general-purposeDefault system prompt for general-purpose worker agents
default-researcheragent_type: researcherDefault system prompt for researcher worker agents
default-revieweragent_type: reviewerDefault system prompt for reviewer worker agents
default-testeragent_type: testerDefault system prompt for tester worker agents
edge-case-researcheragent_type: explorerEdge cases and failure modes researcher - investigates what could go wrong
fleet-architectagent_type: reviewerArchitecture reviewer specialized for agents-fleet coordinator-worker patterns
fleet-ci-engineeragent_type: coderInfrastructure/CI engineer — builds & maintains CI pipelines (test, coverage, regression, build gates)
fleet-ci-revieweragent_type: reviewerCI/build-integrity reviewer — ensures tests, coverage, and green builds are always in place
fleet-coderagent_type: coderCoder agent specialized for agents-fleet (TypeScript/ESM/Ink/Copilot SDK)
fleet-correctness-revieweragent_type: reviewerCorrectness reviewer specialized for agents-fleet type safety and async patterns
fleet-exploreragent_type: explorerRead-only explorer specialized for investigating the agents-fleet codebase
fleet-maintainability-revieweragent_type: reviewerMaintainability reviewer specialized for agents-fleet code hygiene and conventions
fleet-minimalism-revieweragent_type: reviewerMinimalism / over-engineering reviewer — finds what to delete, replace with stdlib/native, or shrink
fleet-performance-revieweragent_type: reviewerPerformance reviewer specialized for agents-fleet runtime efficiency and resource usage
fleet-report-writeragent_type: general-purposeReport-writing specialist — produces polished, interactive HTML reports (agents-fleet light brand theme) with SVG/PNG/interactive diagrams and on-demand PDF export
fleet-researcheragent_type: researcherResearch and investigation agent with scoped write access for the agents-fleet codebase
fleet-security-revieweragent_type: reviewerSecurity reviewer specialized for agents-fleet threat model
fleet-testeragent_type: testerTester agent specialized for agents-fleet (Vitest/better-sqlite3/real implementations)
fleet-workflow-architectagent_type: coderExpert workflow author — composes the workflow-authoring skill, always validates via /wf validate, discovers reusable components
init-api-exploreragent_type: coderInvestigates API surface and contracts
init-arch-exploreragent_type: coderInvestigates architecture and project structure
init-build-exploreragent_type: coderInvestigates build system, test framework, and dev workflow
init-criticagent_type: general-purposeAdversarial critic that verifies investigation findings
init-decision-exploreragent_type: coderInvestigates key design decisions and rationale
init-dep-exploreragent_type: coderInvestigates dependencies and external services
init-pattern-exploreragent_type: coderInvestigates coding patterns and conventions
init-security-exploreragent_type: coderInvestigates security model and authentication
init-synthesizeragent_type: general-purposeSynthesis writer that produces the final /init completion artifacts
inquisitor-edge-casesagent_type: explorerAdversarial edge-cases-angle inquisitor that grills the user with questions about failure modes, security, malformed input, and adversarial scenarios before a feature is built. Spawned by /brainstorm in round 1 alongside inquisitor-ux and inquisitor-technical.
inquisitor-technicalagent_type: explorerAdversarial technical/integration-angle inquisitor that grills the user with questions about architecture, integration points, data shapes, and operational concerns before a feature is built. Spawned by /brainstorm in round 1 alongside inquisitor-ux and inquisitor-edge-cases.
inquisitor-uxagent_type: explorerAdversarial UX-angle inquisitor that grills the user with questions about intent, workflows, success criteria, and user experience before a feature is built. Spawned by /brainstorm in round 1 alongside inquisitor-technical and inquisitor-edge-cases.
performance-revieweragent_type: reviewerPerformance and efficiency focused reviewer
planning-coordinatoragent_type: coordinatorCoordinator specialized for planning workflows - brainstorming, requirements, and design
ponytail-coderagent_type: coderMinimalist coder — a lazy senior developer who writes the least code that works
primary-researcheragent_type: explorerPrimary research perspective - direct investigation of facts, code, documentation, examples
research-coordinatoragent_type: coordinatorCoordinator specialized for parallel research and investigation workflows
research-criticagent_type: reviewerAdversarial critic that cross-references and verifies research findings
sec-binary-analystagent_type: general-purposeStatic reverse engineer for a target binary — triages the file, maps parsing/decoder attack surface, patch-diffs patched-vs-unpatched builds, and produces an RE findings artifact plus a ranked list of fuzz targets
sec-crash-triage-analystagent_type: general-purposeTriages fuzzing crashers and assesses realistic exploitability — reproduces, classifies (!exploitable / !analyze / ASan), determines the memory-safety primitive, weighs mitigations, writes the minimal impact-proving PoC (never weaponized), and produces a triage/exploitability artifact
sec-exploit-poc-engineeragent_type: coderWrites a local, sandboxed failing test that proves a vulnerability in the operator's own code
sec-fixeragent_type: coderImplements the minimal, correct security fix for a finding while preserving behavior
sec-fuzzing-engineeragent_type: general-purposeBuilds and runs a coverage-guided fuzz harness for ONE target routine inside an isolated, disposable Azure VM — assembles seeds + dictionaries, fuzzes under sanitizers, measures coverage, minimizes crashers, and produces a fuzzing artifact
sec-msrc-reporteragent_type: general-purposeProduces the MSRC coordinated-disclosure report from a validated finding plus isolated-VM evidence, using the sec-msrc-report format; WRITES the report artifact
sec-recon-mapperagent_type: general-purposeMaps a codebase's attack surface and writes a structured recon artifact
sec-reporteragent_type: general-purposeWrites the final security report with per-finding detail and an executive summary
sec-triage-analystagent_type: general-purposeDedupes and prioritizes hunter findings and writes a triage summary + fix-target list
sec-verifieragent_type: testerRe-runs the PoC and relevant suites to confirm a fix resolves the finding with no regressions
sec-vm-validatoragent_type: general-purposeOwns the Azure isolated-VM lifecycle — provision an isolated disposable VM, detonate the PoC, capture + hash evidence, assert the security impact, and ALWAYS tear down; uses az + powershell under hard isolation guardrails
sec-vuln-hunteragent_type: reviewerSenior appsec auditor that hunts one vulnerability class and returns evidence-backed findings
security-revieweragent_type: reviewerOWASP-focused security audit reviewer
test-revieweragent_type: reviewerTest coverage and quality reviewer for agents-fleet (TypeScript/Vitest)

Crews

NameWorkflowDescription
agents-fleet-crew/planningDefault development crew for agents-fleet. Composes ponytail-minimalism on fleet-coder and adds a fleet-minimalism-reviewer member.
code-review-crew/code-reviewMulti-perspective code review crew (4 reviewers + synthesis)
init-crew/init-investigationCodebase investigation crew for /init command
minimal-dev-crewMinimalist dev crew — ponytail coder implements, reviewer audits for over-engineering, CI engineer builds pipelines, CI reviewer guards green builds, report-writer produces interactive HTML/PDF reports
research-crew/adversarial-researchAdversarial research crew — 4 angles + critic
security-crew/security-autopilotApplication-security audit crew — recon → hunt → triage → PoC → fix → verify → report
workflow-authoring-crew/freeformAuthoring crew — a workflow architect composes the workflow-authoring skill to design, validate, and save new .workflow.md files.