|
1 | | -<div align="center"> |
2 | | - |
3 | | -# Basilisk Labs |
4 | | - |
5 | | -**Infrastructure for long-range collaboration between people and autonomous agents.** |
6 | | - |
7 | | -<p> |
8 | | - <a href="https://basilisk-labs.com"><img alt="Website" src="https://img.shields.io/badge/website-basilisk--labs.com-111827?style=for-the-badge"></a> |
9 | | - <a href="https://agentplane.org"><img alt="AgentPlane" src="https://img.shields.io/badge/project-AgentPlane-2563eb?style=for-the-badge"></a> |
10 | | - <a href="mailto:hello@basilisk-labs.com"><img alt="Email" src="https://img.shields.io/badge/contact-hello%40basilisk--labs.com-0f766e?style=for-the-badge"></a> |
11 | | -</p> |
12 | | - |
13 | | -<p> |
14 | | - <img alt="Git-native" src="https://img.shields.io/badge/Git--native-111827?style=flat"> |
15 | | - <img alt="Agent work" src="https://img.shields.io/badge/agent_work-evidence--first-2563eb?style=flat"> |
16 | | - <img alt="Audit trail" src="https://img.shields.io/badge/audit_trail-reviewable-0f766e?style=flat"> |
17 | | - <img alt="Open source" src="https://img.shields.io/badge/open_source-practical-7c3aed?style=flat"> |
18 | | -</p> |
19 | | - |
20 | | -</div> |
21 | | - |
22 | | -Basilisk Labs builds tools for agentic software work that needs memory, coordination, and |
23 | | -accountability over time. Our current center of gravity is **AgentPlane**: an open-source audit |
24 | | -layer that turns coding-agent sessions into reviewable Git artifacts. |
25 | | - |
26 | | -We are interested in systems where the important question is not just whether an agent produced a |
27 | | -diff, but whether a human or another agent can reconstruct the task, plan, approval, verification, |
28 | | -and closure chain later. |
29 | | - |
30 | | -## Focus |
31 | | - |
32 | | -| Area | What we build | |
33 | | -| --- | --- | |
34 | | -| Agent coordination | Repo-local workflows for planning, handoff, verification, and finish states. | |
35 | | -| Change evidence | Agent Change Records, task artifacts, signed catalogs, and reviewable traces. | |
36 | | -| Developer tooling | CLI-first tools that work with Codex, Claude Code, Cursor, Aider, and GitHub Actions. | |
37 | | -| Long-lived memory | Practices and infrastructure that preserve useful context without depending on chat history. | |
38 | | - |
39 | | -## Projects |
40 | | - |
41 | | -| Project | Status | Description | |
42 | | -| --- | --- | --- | |
43 | | -| [agentplane](https://github.com/basilisk-labs/agentplane) | Open source | Audit layer for coding agents. Keeps task, plan, verification, and closure evidence in Git. | |
44 | | -| [agent-change-record](https://github.com/basilisk-labs/agent-change-record) | Open standard | JSON schema and tooling for machine-readable evidence manifests for agent-authored changes. | |
45 | | -| [agentplane-recipes](https://github.com/basilisk-labs/agentplane-recipes) | Catalog | Signed recipes for reusable AgentPlane workflows and development practices. | |
46 | | -| [setup-agentplane](https://github.com/basilisk-labs/setup-agentplane) | CI | GitHub Action for installing AgentPlane inside CI runners. | |
47 | | -| [homebrew-tap](https://github.com/basilisk-labs/homebrew-tap) | Distribution | Homebrew tap for AgentPlane CLI releases. | |
48 | | -| [scoop-bucket](https://github.com/basilisk-labs/scoop-bucket) | Distribution | Scoop manifests for AgentPlane on Windows. | |
49 | | -| [agentplane-harbor-adapter](https://github.com/basilisk-labs/agentplane-harbor-adapter) | Experimental | Benchmark adapter for evaluating AgentPlane as a control-plane wrapper around coding agents. | |
50 | | -| [agentplane-marketing](https://github.com/basilisk-labs/agentplane-marketing) | Knowledge base | Public positioning, editorial, launch, and market-practice source documents for AgentPlane. | |
51 | | - |
52 | | -## Principles |
53 | | - |
54 | | -- Keep workflow evidence close to the code. |
55 | | -- Prefer explicit plans and verification over invisible agent state. |
56 | | -- Make artifacts useful to both humans and machines. |
57 | | -- Treat benchmarks, recipes, and integrations as evidence surfaces, not just growth assets. |
58 | | -- Build small, inspectable tools before adding hosted complexity. |
59 | | - |
60 | | -## Start here |
61 | | - |
62 | | -```bash |
63 | | -npm i -g agentplane |
64 | | -agentplane init |
65 | | -agentplane quickstart |
66 | | -``` |
67 | | - |
68 | | -Read the docs at [agentplane.org](https://agentplane.org) or start from the |
69 | | -[AgentPlane repository](https://github.com/basilisk-labs/agentplane). |
70 | | - |
71 | | -## Contact |
72 | | - |
73 | | -- Email: [hello@basilisk-labs.com](mailto:hello@basilisk-labs.com) |
74 | | -- Website: [basilisk-labs.com](https://basilisk-labs.com) |
75 | | -- AgentPlane: [agentplane.org](https://agentplane.org) |
| 1 | +<p align="center"> |
| 2 | + <img src="https://raw.githubusercontent.com/basilisk-labs/.github/main/profile/assets/basilisk-labs-header.svg" alt="Basilisk Labs GitHub profile header" style="width:100%;max-width:100%;"> |
| 3 | +</p> |
| 4 | + |
| 5 | +<div align="center"> |
| 6 | + |
| 7 | +# Basilisk Labs |
| 8 | + |
| 9 | +**Local-first infrastructure for inspectable autonomous-agent work.** |
| 10 | + |
| 11 | +<p> |
| 12 | + <a href="https://basilisk-labs.com"><img alt="Website" src="https://img.shields.io/badge/website-basilisk--labs.com-111827?style=for-the-badge"></a> |
| 13 | + <a href="https://agentplane.org"><img alt="AgentPlane" src="https://img.shields.io/badge/project-AgentPlane-ff4d00?style=for-the-badge"></a> |
| 14 | + <a href="https://github.com/basilisk-labs/agentplane/releases/tag/v0.6.2"><img alt="AgentPlane v0.6.2" src="https://img.shields.io/badge/AgentPlane-v0.6.2-18794e?style=for-the-badge"></a> |
| 15 | + <a href="mailto:hello@basilisk-labs.com"><img alt="Email" src="https://img.shields.io/badge/contact-hello%40basilisk--labs.com-0f766e?style=for-the-badge"></a> |
| 16 | +</p> |
| 17 | + |
| 18 | +<p> |
| 19 | + <img alt="Git-native" src="https://img.shields.io/badge/Git--native-evidence-111827?style=flat"> |
| 20 | + <img alt="Local-first" src="https://img.shields.io/badge/local--first-no_hosted_runtime-2563eb?style=flat"> |
| 21 | + <img alt="Agent Change Record" src="https://img.shields.io/badge/ACR-reviewable_manifest-7ea2ff?style=flat"> |
| 22 | + <img alt="Recipes" src="https://img.shields.io/badge/recipes-signed_catalog-6e56cf?style=flat"> |
| 23 | +</p> |
| 24 | + |
| 25 | +</div> |
| 26 | + |
| 27 | +Basilisk Labs builds open-source systems for agentic software work that must remain reconstructable |
| 28 | +after the terminal session is gone. The center of gravity is **AgentPlane**: a Git-native workflow |
| 29 | +control layer that records task intent, accepted plans, verification evidence, finish state, and |
| 30 | +Agent Change Records beside the code. |
| 31 | + |
| 32 | +The operating premise is simple: a pull request should show more than a diff. It should preserve |
| 33 | +why the change exists, which policy bounded it, which checks ran, and what another human or agent |
| 34 | +can audit later. |
| 35 | + |
| 36 | +## Current work |
| 37 | + |
| 38 | +| Surface | Repository | What it is | |
| 39 | +| --- | --- | --- | |
| 40 | +| Workflow control | [agentplane](https://github.com/basilisk-labs/agentplane) | CLI for task -> plan -> approve -> implement -> verify -> finish, with repo-local evidence and ACR. | |
| 41 | +| Evidence standard | [agent-change-record](https://github.com/basilisk-labs/agent-change-record) | JSON schema and tooling for machine-readable manifests of agent-authored changes. | |
| 42 | +| Reusable behavior | [agentplane-recipes](https://github.com/basilisk-labs/agentplane-recipes) | Signed recipe catalog for reusable coding-agent workflows, repo mapping, and development playbooks. | |
| 43 | +| Blueprint catalog | [agentplane-blueprints](https://github.com/basilisk-labs/agentplane-blueprints) | Installable AgentPlane blueprints for verification routes and task execution patterns. | |
| 44 | +| CI install path | [setup-agentplane](https://github.com/basilisk-labs/setup-agentplane) | GitHub Action for installing AgentPlane in CI runners from release assets. | |
| 45 | +| Release channels | [homebrew-tap](https://github.com/basilisk-labs/homebrew-tap), [scoop-bucket](https://github.com/basilisk-labs/scoop-bucket) | Homebrew and Scoop distribution for the AgentPlane CLI. | |
| 46 | +| Evaluation adapters | [agentplane-harbor-adapter](https://github.com/basilisk-labs/agentplane-harbor-adapter), [agentplane-bitgn-adapter](https://github.com/basilisk-labs/agentplane-bitgn-adapter) | Benchmark adapters for running AgentPlane as a reproducible coding-agent harness. | |
| 47 | +| Research workspace | [openclaw-deus](https://github.com/basilisk-labs/openclaw-deus) | Experimental automation workspace for local agent workflow research and prototypes. | |
| 48 | + |
| 49 | +## AgentPlane in one screen |
| 50 | + |
| 51 | +```text |
| 52 | +task -> plan -> approve -> implement -> verify -> finish |
| 53 | + | |
| 54 | + v |
| 55 | + Agent Change Record |
| 56 | +``` |
| 57 | + |
| 58 | +AgentPlane is not another coding agent. Claude Code, Codex, Cursor, Aider, and similar tools do the |
| 59 | +editing. AgentPlane wraps that work in local lifecycle state, policy routing, verification notes, |
| 60 | +and reviewable artifacts. |
| 61 | + |
| 62 | +```bash |
| 63 | +npm i -g agentplane |
| 64 | +agentplane init |
| 65 | +agentplane quickstart |
| 66 | +``` |
| 67 | + |
| 68 | +Core surfaces: |
| 69 | + |
| 70 | +- **ACR:** commit-safe JSON evidence for task intent, approved plan, commits, verification, and merge readiness. |
| 71 | +- **Local context:** repository-owned memory through `context/raw`, `context/wiki`, sourced facts, graph data, and derived projections. |
| 72 | +- **Workflow modes:** `direct` for local loops, `branch_pr` for task branches, worktrees, PR review artifacts, and integration flow. |
| 73 | +- **Recipes and blueprints:** signed reusable behavior modules for repo mapping, agent profiles, prompts, and verification paths. |
| 74 | +- **Distribution:** npm, Homebrew, Scoop, standalone release assets, and GitHub Actions setup. |
| 75 | + |
| 76 | +## Principles |
| 77 | + |
| 78 | +- Keep workflow evidence close to the code. |
| 79 | +- Prefer explicit plans and verification over invisible agent state. |
| 80 | +- Preserve local ownership: no hosted runtime is required for the core workflow. |
| 81 | +- Make artifacts useful to both humans and machines. |
| 82 | +- Treat benchmarks, recipes, and integrations as evidence surfaces, not just growth assets. |
| 83 | + |
| 84 | +## Links |
| 85 | + |
| 86 | +| Destination | Link | |
| 87 | +| --- | --- | |
| 88 | +| Website | [basilisk-labs.com](https://basilisk-labs.com) | |
| 89 | +| AgentPlane docs | [agentplane.org](https://agentplane.org) | |
| 90 | +| AgentPlane repository | [github.com/basilisk-labs/agentplane](https://github.com/basilisk-labs/agentplane) | |
| 91 | +| Contact | [hello@basilisk-labs.com](mailto:hello@basilisk-labs.com) | |
| 92 | + |
| 93 | +<sub>The header is generated by `node scripts/generate-profile-header.mjs` and checked with `node scripts/generate-profile-header.mjs --check`.</sub> |
0 commit comments