🤖 Bots and examples
Iterion ships a maintained fleet of production bots — a co-CTO, a feature builder, security auditors, a dependency upgrader, and more — each a declarative .bot you can run, inspect, or fork. The repository keeps two things deliberately separate:
bots/contains the maintained bot catalogue. Each bot is a folder with amain.bot, amanifest.yaml, and any skills or resources it needs.examples/contains focused DSL and integration demos. They are teaching fixtures, not the product bot catalogue.
Iterion runs .bot sources and packaged .botz bundles. Other workflow extensions are rejected consistently at the CLI, server, dispatcher, and studio boundaries; pkg/dsl/workflowfile is the source of truth.
Embedded dispatcher catalogue
The release binary embeds these nine general-purpose bots for zero-config dispatcher use. The checked-in folders remain the editable sources; task templates:dispatch-bots refreshes the embedded copies.
| Persona | Bot | Purpose |
|---|---|---|
| 🧭 Nexie | whats-next | Conversational co-CTO: inspect the repo and board, recommend work, curate tickets, and dispatch the right bot. |
| 🛠️ Featurly | feature-dev | Deliver one feature through an adaptive implementation campaign and deterministic verification gates. |
| 🌿 Billy | branch-improve-loop | Review and improve the branch diff, committing verified fixes in stride until convergence. |
| 🌍 Willy | whole-improve-loop | Apply one cross-cutting improvement axis across an existing codebase, site by site. |
| 📚 Doki | docs-refresh | Align documentation with code using one adaptive campaign, advisory drift hints, and a deterministic Markdown-only scope gate; never edits code. |
| 🔎 Revi | review-pr | Read-only pull/merge-request review with grounded findings and forge/board publication. |
| 🛡️ Seki | sec-audit-source | Read-only source security audit using SAST, secret scanning, triage, and false-positive memory. |
| 📦 Depsy | sec-audit-deps | Read-only dependency malware/CVE audit with a cross-run package cache. |
| ⬆️ Renovacy | secured-renovacy | Multi-stack dependency upgrades with security, compatibility, build, and review gates. |
iterion dispatch also exposes a default alias. Run iterion bots list to inspect the catalogue visible from the current workspace.
Additional maintained bots
These bots are shipped in the repository and are discoverable by the CLI and studio when bots/ is in the search path, but they are not copied into the zero-config dispatcher template:
| Persona | Bot | Purpose |
|---|---|---|
| 🗺️ Adry | adr-cartograph | Reconcile implemented architectural decisions with docs/adr/ and surface decision gaps. |
| ⚖️ ReArchi | adr-rechallenge | Human-guided re-evaluation of an existing ADR: keep, change, or append an addendum. |
| 🏗️ Appy | app-dev | Build a new application end to end, optionally beginning with a conversational specification. |
| 🎭 Bmady | bmady | Human-steered analyst → PM → architect → developer → QA delivery pipeline. |
| 💂 Vetty | dep-update-guard | Audit and align automated dependency-update PRs without merging them. |
| 🧰 Devy | devbox-setup | Create and validate a pinned devbox.json/lock for a reproducible project toolchain. |
| 🧬 Evoly | evolve | Long-horizon product and architecture partner backed by persistent per-bot memory. |
| 🧩 Fini | feature-gap-fill | Finish a structured partial-implementation gap while preserving what already works. |
| 🔭 Vigie | feed-watch | Collect RSS/Atom feeds at zero LLM cost, then produce and deliver grounded digests. |
| 🏷️ Triagy | issue-triage | Single-shot card triage: read a fresh board card, classify it, stamp the handler bot + labels, and leave a routing comment. Routes work to other bots; never dispatched to. |
| 💬 Revi (converse) | revi-converse | Answer a focused /revi follow-up in the same forge discussion; never edits code. |
| ♿ Acci | rgaa-audit | Read-only RGAA 4.1.2 source audit with a deterministic coverage gate. |
| ⛓️ Shieldy | supply-shield | Diff-scoped dependency-malware gate for forge events. |
| 🚨 Vulny | supply-shield-cve | Diff-scoped known-CVE gate for changed dependency versions. |
| 🧪 Testy | test-coverage | Add meaningful regression-catching tests and verify both the suite and the new test diff. |
| 📖 Wikky | wiki-gen | Generate and incrementally maintain a validated Open Knowledge Format wiki. |
| 🪞 Goldy | golden-master | Build a behavioural non-regression net (golden master) for an existing app and prove it is neither blind nor hysterical with a deterministic mutation counter-test; the safety net under a migration or refactor. |
| 🧱 Morphy | modernize | Carry a repo through modernisation lots (toolchain/runtime/framework/datastore) one deterministic gate-to-gate step at a time, each proven by a behavioural oracle — refuses to start without a golden-master net. |
The manifests are authoritative for inputs, invocation modes, required capabilities, forge events, and suggested schedules. Nexie’s generated bot decision catalogue is the routing-oriented view.
Focused DSL demos
| Example | Shows |
|---|---|
human-in-the-loop.bot | A human entry node and interaction form. |
human-file-upload.bot | A file-typed gate field: the operator drops a file in the run console (or --answer track=@./x.mp3) and a tool node opens the real bytes. No LLM. |
async-questions/ | interaction: async — non-blocking ask_user_async questions with an await_answers sync point (ADR-081). |
clarify/ | Conversational read-only facilitation with LLM interaction. |
composition/ | group/use, nested bots, and parallel sub-bot execution. |
cursors/ | Cursor declarations and per-node calibration. |
events/ | In-run emit/wait coordination. |
turing/ | Fuelled expression/loop semantics. |
supervisor/ | Concurrent supervisor declarations and steering. |
ultracode/ | Ultracode compression mode. |
web-search/ | Tiered web-search capability use. |
keepalive/ | Sub-minute always-on schedule shape and overlap policy. |
nested-subbots-demo/ | Multi-level child-run nesting. |
pipeline-board-demo/ | Pipeline-board episode projection. |
devcontainer-devbox/ | Devcontainer sandbox plus repository devbox.json, including non-interactive PATH provisioning and composition with bot-local tools. |
github-actions/ | Human-in-the-loop workflow from GitHub Actions. |
Other root-level fixtures exercise explicit else, deploy E2E, and the review/merge gate. Validate any example with iterion validate <path> before adapting it.