Disclosure: Software Sausage is our product. Agency Agents did not sponsor, review, or endorse this article. AI tools helped draft and edit it; the evidence boundary is stated below.
The Agency Agents repository is difficult to ignore: 264 specialized agent definitions, broad coding-harness support, and—when I reviewed it on September 9, 2026—roughly 151,000 GitHub stars.
The tempting conclusion is that a larger virtual team produces better work.
The repository does not establish that. What it does provide is a useful, MIT-licensed role library and a competent installer. The value comes from selecting a few narrow roles and making their outputs pass real checks.
This is a source review at commit 6d29a9b, not a benchmark. I inspected the role files, installer, converter, contribution rules, workflow example, and current GitHub checks. I did not evaluate every integration or the desktop app.
Agent definitions are not one-line personas. They name deliverables, workflows, constraints, and success metrics. The project converts them for Claude Code, Codex, Cursor, Gemini CLI, OpenCode, Qwen Code, Aider, and other harnesses.
More importantly, the installer lets you select one role or division, show a dry run, and target an explicit path. Agency Agents itself warns that OpenCode currently registers only about 119 agents and recommends installing a subset.
The project's contribution guide contains the best design rule: a new agent needs a narrow specialization, distinct behavior, concrete deliverables, measurable success, and real testing. Near-duplicate re-skins are rejected.
That rule should apply to the workflow too. If two roles produce the same artifact, if nobody consumes an output, or if success cannot be checked, remove the role.
The README's “never sleep” and “always deliver” language is marketing. Prompts still fail, time out, overrun context, and agree on the same plausible mistake.
A role prompt also does not create process isolation. Giving a “reviewer” the same credentials and write access as the implementer makes the label cosmetic.
The repository's seven-agent startup example recommends passing full outputs between roles. Its newer Multi-Agent Systems Architect warns about context growth and recommends summaries and structured state. Use the second rule.
Green installer and manifest checks show that the repository is maintained and internally consistent. They do not show that seven agents outperform one capable agent on correctness, elapsed time, or cost.
For a software change, three responsibilities are enough: The implementer gets at most two corrections. Then a person resolves the findings, reruns the repository checks, captures fresh runtime proof, and decides whether to ship.
The free [three-agent code-change kit](https://softwaresausage.com/ai/recipes/three-agent-code-change-gate?source=community&utm_campaign=dev_agency_agents) includes the handoff ledger and verifier.
If the question is whether orchestration helps, compare it with one agent:
The free [multi-agent experiment kit](https://softwaresausage.com/ai/recipes/multi-agent-experiment-gate?source=community&utm_campaign=dev_agency_agents) provides the comparison manifest.
Start with one agent. Add one role only when it owns a distinct artifact with a consumer and an acceptance check. Keep permissions separate. Cap retries. Leave the human at the irreversible boundary.
Agency Agents makes roles easy to install. The harder—and more valuable—work is proving which ones deserve to stay.
The complete source review and both pullable kits are at Software Sausage. The public Software Sausage MCP endpoint can also find these recipes and return their complete run ledgers inside a compatible agent client.