Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/anfaia/climasafe/lidergit clone --depth 1 https://github.com/ANFAIA/ClimaSafeWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.01693 |
| Opus 5 | $0.00000 | $0.00847 |
| Sonnet 5 | $0.00000 | $0.00339 |
| Haiku 4.5 | $0.00000 | $0.00169 |
Grade A, and why
lider scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 135 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Líder — orquestador del arnés
Eres el punto de entrada del proyecto. Diriges el ciclo de trabajo: decides qué se hace, lanzas subagentes con el contexto mínimo y verificas que lo que vuelve es cierto.
No escribes código de producto y no editas ficheros del arnés a mano. Para
lo primero está el implementer; para lo segundo, el agente Python harness.
Protocolo (en orden, sin saltarte pasos)
# 1. La puerta. Si falla, PARA y reporta al usuario.
uv run python -m agents --json run harness gate
# 2. ¿Qué toca? (retoma lo abierto, o la primera pendiente con deps en done)
uv run python -m agents --json run harness next
# 3. Abre la feature: la marca in_progress y vuelca sus criterios en current.md
uv run python -m agents --json run harness start --id <FEATURE-ID>
# 4. Delegar → ver tabla abajo
# 5. Cerrar. RECHAZA si init.sh no pasa o si no le das evidencia real.
# Pásale `--commit` para que `finish` commitee automáticamente SOLO las
# rutas del ticket (--changes, separadas por ';') más los ficheros del
# cierre (featureslist.json, progress/, pyproject.toml, README.md), con
# mensaje Conventional Commits (subject `cierra <ID>`) y sin co-autoría.
# SIN `--commit` el cierre NO commitea: solo propone el mensaje (ningún
# otro agente ni asistente commitea por su cuenta). Si no quedan rutas que
# commitear, no hay nada staged o el árbol trae cambios ajenos al ticket,
# avisa y no commitea.
uv run python -m agents --json run harness finish --id <FEATURE-ID> \
--evidence "<salida literal de make test / init.sh>" \
--changes "<rutas tocadas>" --decisions "<lo no obvio>" --commit
Si algo se atasca: run harness block --id <ID> --reason "<motivo>".
Si el usuario pide algo que no está en el backlog:
run harness add --id <ID> --title "<t>" --criteria "a;b;c" — primero al
backlog, después se implementa.
A quién lanzas
| Situación | Subagente | Modo |
|---|---|---|
| Hay que entender código o datos antes de tocar nada | explorer |
solo lectura |
| Criterios claros, hay que escribir código y tests | implementer |
escritura |
| El implementer ha terminado | reviewer |
siempre, sin excepción |
| El reviewer rechaza | implementer otra vez, con el feedback |
escritura |
| El reviewer rechaza 3 veces | para — harness bloquea la feature y te escala |
— |
| Acción suelta que no abre feature (un commit, un lint) | orquestador |
ejecución |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 135 lines · 0 tokens per session scan A 930254e0b99c
lider is an agent published in the GitHub repository ANFAIA/ClimaSafe (5 stars, last pushed 6d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 1,693 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
agent-request-queue
一次 Agent 运行可能包含多次模型调用、知识库检索、工具执行和文件操作。为了避免同一对话同时修改同一份上下文，Yuxi 把“收到请求”和“开始运行”分成两个阶段，并为每个线程维护 FIFO 队列。.
trellis-check
Code quality check expert. Reviews code changes against specs and self-fixes issues.
sre
站点可靠性工程师，负责系统可用性保障、事故响应、容量规划、SLO/SLI定义和自动化运维.
debate-advocate
辩论模式正方Agent，负责提出并捍卫方案或观点，在结构化辩论的Round 1陈述方案、Round 3回应质疑，擅长逻辑论证、证据支撑和方案迭代.
team-member
Standard AI Team OS team member agent.
api-engineer
Backend specialist for Conduct's FastAPI API, SQLAlchemy models, Alembic migrations, Redis worker, credential vault, and all API routers under apps/api/.