10 KiB
dev_agent_team
A distributable package of 14 opencode agent definitions plus a one-command installer, so the same agent team can be set up identically on any machine.
What this is
This repository packages a complete multi-agent team for opencode — 14 role-specialized agents that work as one system:
architect, builder, breakdowner, designer, detective, explorer,
maintainer, orchestrator (primary), philosopher, reviewer, tester,
toolsmith, workflow-architect, writer.
The agent definitions live in agents/ and are copied verbatim into your
opencode config directory by scripts/install.sh. The installer is idempotent:
re-running it is safe, and it backs up any pre-existing files it would
overwrite into a timestamped folder under <target>/.backup/, keeping only
the 5 most recent backup folders.
Repository Intelligence Bootstrap
The team includes a Repository Intelligence Bootstrap system: when the
Orchestrator starts work on an unfamiliar repository, it automatically analyzes
the repository and generates a local .opencode/ knowledge layer
(context, architecture, build/test, conventions, deployment). Every agent then
reads this pre-analyzed context instead of re-discovering repository fundamentals.
The bootstrap is idempotent, language-agnostic, and preserves manually enriched content. See docs/REPOSITORY_INTELLIGENCE.md for the full architecture.
Adaptive, Evidence-Driven Orchestration
The team is architected as a decision engine: the Orchestrator recalls project memory, understands the task, estimates complexity, loads repository intelligence, chooses the next best action from an explicit action catalog, verifies outcomes independently, re-plans when evidence changes, stores durable findings, and stops when sufficiently verified. Every agent produces structured evidence-state records and knows when to stop and escalate.
See docs/AGENT_ARCHITECTURE.md for the full architecture, and docs/EVALUATION_SCENARIOS.md for the 12 runtime evaluation scenarios used to measure the team's own quality.
Project Memory & Skills
The team includes cross-session project memory (memory/) for persisting
decisions, lessons, failures, architecture notes, and session state. A
lifecycle script (scripts/memory-lifecycle.sh) provides deterministic recall,
store, list, search, and cleanup operations.
The team also includes 12 reusable skills (skills/) — specialized
methodologies for TDD, debugging, architecture, code review, security review,
and more. Agents load relevant skills when dispatched.
Both systems require human approval for changes to core agent behavior
(improvements/).
Repository layout
dev_agent_team/
├── README.md # this file
├── .gitignore
├── agents/ # the 14 agent definitions (*.md)
├── .tasks/ # per-project Task Breakdown (gitignored; created by breakdowner, never committed)
├── memory/ # cross-session project memory
│ ├── MEMORY.md # index with lifecycle rules
│ ├── decisions/ # architectural/technical choices
│ ├── lessons/ # reusable knowledge
│ ├── failures/ # root causes + prevention
│ ├── architecture/ # system structure documentation
│ └── sessions/ # work-in-progress state
├── skills/ # 12 reusable specialized methodologies
│ ├── SKILLS.md # index with loading rules
│ ├── tdd/SKILL.md # Test-Driven Development
│ ├── systematic-debugging/SKILL.md # debugging methodology
│ ├── architecture-design/SKILL.md # architecture decisions
│ ├── code-review/SKILL.md # code review process
│ ├── security-review/SKILL.md # security review
│ ├── repository-analysis/SKILL.md # repo exploration
│ ├── failure-analysis/SKILL.md # failure investigation
│ ├── refactoring/SKILL.md # safe code restructuring
│ ├── test-analysis/SKILL.md # test quality assessment
│ ├── incident-investigation/SKILL.md # production incidents
│ ├── browser-automation/SKILL.md # web interaction patterns
│ └── research/SKILL.md # information gathering
├── improvements/ # proposal-based improvement system
│ ├── README.md # proposal format and lifecycle
│ ├── pending/ # proposals awaiting approval
│ ├── applied/ # approved and implemented
│ └── rejected/ # not approved
├── scripts/
│ ├── install.sh # one-command installer (agents + runtime tree; --uninstall/--migrate/--self-test)
│ ├── repo-bootstrap.sh # repository intelligence bootstrap tool
│ ├── memory-lifecycle.sh # memory CRUD operations
│ ├── test-all.sh # single entry point that runs all 8 test suites
│ ├── test-repo-bootstrap.sh # test suite for the bootstrap
│ ├── test-agent-architecture.sh # structural tests for the agent architecture
│ ├── test-memory-system.sh # structural tests for memory/skills/improvements
│ ├── test-integration.sh # end-to-end memory/skills/lifecycle integration tests
│ ├── test-install.sh # runtime-tree install + idempotency tests
│ ├── test-runtime.sh # runtime-after-source-deletion + bootstrap tests
│ ├── test-path-resolution.sh # prompt/skills/improvements path-resolution tests
│ ├── test-memory-isolation.sh # per-project memory isolation tests
│ └── verify-permission-patterns.sh # permission engine verifier
└── docs/
├── PROMPT_INSTALL.md # paste-ready prompt for installing from inside opencode
├── REPOSITORY_INTELLIGENCE.md # bootstrap architecture documentation
├── AGENT_ARCHITECTURE.md # adaptive/evidence-driven architecture documentation
└── EVALUATION_SCENARIOS.md # runtime evaluation scenarios for the team itself
Quickstart
git clone git@gitea.skink-platy.ts.net:admin/dev_agent_team.git
cd dev_agent_team
./scripts/install.sh
To install somewhere other than the default location:
OPENCODE_AGENTS_DIR=/path/to/opencode/agents ./scripts/install.sh
The installer also installs the self-contained runtime tree (scripts bin/,
12 general-purpose skills skills/, improvements store, and an
install-manifest.json) under $OPENCODE_DEV_AGENT_TEAM (default
$XDG_CONFIG_HOME/opencode/dev-agent-team, fallback
~/.config/opencode/dev-agent-team). After installation the source checkout may
be deleted; the agents and runtime keep working from any project.
Runtime-aware operations:
./scripts/install.sh # install agents + full runtime tree (idempotent)
./scripts/install.sh --uninstall # remove manifest-tracked agents + runtime (keeps backups; preserves improvements/ unless --purge)
./scripts/install.sh --uninstall --purge # also remove improvements/ user data
./scripts/install.sh --migrate # one-time import of pending improvement proposals (no-op today)
./scripts/install.sh --self-test # run runtime test suites in a throwaway temp HOME
--uninstall never touches project memory/, .opencode/, AgentsReport/,
or ~/.config/opencode/opencode.json.
Repository bootstrap
After installing the agents, the bootstrap tool is available at
scripts/repo-bootstrap.sh. From inside any target repository:
# Check freshness (exit 0=fresh, 1=stale/missing)
repo-bootstrap.sh status
# Create/update .opencode/ knowledge layer
repo-bootstrap.sh bootstrap
# Force regenerate generated files
repo-bootstrap.sh refresh
Run all test suites with a single command (aggregates the eight suites below):
bash scripts/test-all.sh
105 individual checks across 8 suites (17 architecture + 16 memory + 11 bootstrap
- 12 integration + 15 install + 10 runtime + 12 path-resolution + 12 memory-isolation). Any suite failing makes the overall exit code non-zero.
Individual suites:
test-agent-architecture.sh— architecture contains the required adaptive/evidence-driven elements (17 structural tests).test-memory-system.sh— memory, skills, and improvement systems are structurally sound (16 tests).test-repo-bootstrap.sh— bootstrap behavior and idempotency (11 tests covering all 10 acceptance criteria).test-integration.sh— end-to-end memory/skills/improvements integration (12 tests).test-install.sh— runtime-tree install, manifest, rc export, backups, and idempotency (15 tests).test-runtime.sh— runtime survives after source deletion and works from any unrelated project (10 tests).test-path-resolution.sh— prompt/skills/improvements resolve to the runtime root; project-scoped refs stay project-relative (12 tests).test-memory-isolation.sh— per-project memory isolation viaOPENCODE_MEMORY_DIR(12 tests).
Manual install alternative
Prefer to do it yourself? A plain copy achieves the same result:
mkdir -p ~/.config/opencode/agents
cp -p agents/*.md ~/.config/opencode/agents/
(You will not get the automatic backup of pre-existing files that
scripts/install.sh provides.)
Prompt-based install
Already inside an opencode session? Copy a ready-made prompt that instructs the primary agent to clone this repo, run the installer, and verify the result: see docs/PROMPT_INSTALL.md.
Requirements
- opencode installed (needed to actually use the agents)
bashand coreutils (cp,mkdir,date,sha256sumorcksum)- SSH access to the gitea host for cloning (or an HTTPS remote, if mirrored)
Verification
After installing:
-
Restart any running opencode sessions so they pick up the new agents.
-
Run:
opencode agent list -
You should see exactly 14 agents: architect, builder, breakdowner, designer, detective, explorer, maintainer, orchestrator, philosopher, reviewer, tester, toolsmith, workflow-architect, writer.