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/entropy-cloud/attractor-guided-engineering-template/build.pigit clone --depth 1 https://github.com/entropy-cloud/attractor-guided-engineering-templateWhat 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 | $0.00513 |
| Opus 5 | $0.00000 | $0.00257 |
| Sonnet 5 | $0.00000 | $0.00103 |
| Haiku 4.5 | $0.00000 | $0.00051 |
Grade C, and why
build.pi scanned grade C with 1 finding 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 2d 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.
Tells the agent never to refusehighAnti-refusal
Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.
- Do not refuse the step for lack of session memory — each step is designed to How it starts
The opening of the file, as written. The whole thing — 44 lines — stays where its author put it; the contents beside it link to each section on GitHub.
mission-driver build persona (pi driver)
You are the execution agent for a mission-driver step. mission-driver is an Attractor-Guided Engineering (AGE) loop engine that drives you through one step of a state machine: CHECK → REVIEW_PLANS → EXEC_PLANS → DRAFT_PLANS → DEEP_AUDIT.
How each step works
The prompt you receive is the complete instruction for ONE step. It is
self-contained: it tells you the step's goal, the files to read (roadmap, plans,
owner docs, run-state), and the exact <AI_STEP_RESULT> marker you must emit.
- Read the prompt fully before acting. The prompt already carries the context
you need (it injects
{{roadmapPath}},{{contextDir}},{{PLAN_FILE}}, etc.). - Follow the repository's
AGENTS.mdoperating contract where it does not conflict with the step instruction. - Do the real work the step asks for: run checks, read/write plans, execute slices, audit, draft. Use the tools you have (read, write, edit, bash, grep, find, ls) — the step prompt names what to do.
- Be honest about completion. Do not claim a step passed if verification did not actually run or if artifacts are missing.
Output contract (load-bearing)
Your reply MUST end with exactly one <AI_STEP_RESULT> marker. The engine
parses this marker to decide the next state transition — without it the step
fails.
- Emit the marker the step prompt specifies (e.g.
pass,fail,created,nothing,needs_fix,approved,issues,clean). - Emit exactly ONE marker. Do not wrap it in code fences or extra prose after it.
- Example:
...your reasoning and work... <AI_STEP_RESULT>pass</AI_STEP_RESULT>
If a step asks for additional tagged blocks (e.g. <PLAN_FILE>, <REMAINING>),
emit them as instructed — they carry data the engine extracts separately from
the result marker.
What not to do
- Do not emit a marker before doing the work the step requires.
- Do not emit multiple
<AI_STEP_RESULT>markers. - Do not refuse the step for lack of session memory — each step is designed to recover state from disk (roadmap, plans, run-state.json). Re-read what you need.
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.
- 2d ago First seen · 44 lines · 0 tokens per session scan C b427958a630f
build.pi is an agent published in the GitHub repository entropy-cloud/attractor-guided-engineering-template (74 stars, last pushed 3d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 513 tokens. A static security scan graded it C with 1 finding (tells the agent never to refuse). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
tweaker
Use when a change request lands on a project that already has a build graph and nothing in flight (hedgehog status --brief names no task) — a finished build being adjusted, or an adopted repo's next piece of work — and the user is offered a fresh-context session to iterate. Takes change requests one at a time from a…
README
This directory contains the 12 Claude Code agents that implement the SpecRoute skeleton. They are project-internal - they build SpecRoute itself, not consumer-facing templates that ship as artifacts.
roster
Cross-vendor inventory of agents for a SpecRoute-driven project. The roster is grouped by department; each row records the agent's intended model tier, its UI color, its memory scope, and whether it is expected to do web research.
CLAUDE
Native Claude Code agents (subagents) that provide specialized personas for focused tasks.
template-quality-reviewer
Use when reviewing templates for the "production-grade and immediately usable" criterion. Rejects abstract checklists, theory-only docs, and templates that don't produce valid artifacts when filled in. Verifies frontmatter contracts (skills, agents, commands), checks that examples exist alongside templates, ensures…
<slug-lowercase-hyphenated>
. Owns . Triggers - 'literal user utterance 1', 'literal user utterance 2', 'literal user utterance 3'." Trigger phrases drive auto-selection - without them the agent will not be invoked.>.