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 skills add JairoTorregrosa/jaiskills --skill goal-loopgit clone --depth 1 https://github.com/JairoTorregrosa/jaiskillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/jairotorregrosa/jaiskills/goal-loop)<a href="https://agentmods.dev/skills/jairotorregrosa/jaiskills/goal-loop"><img src="https://agentmods.dev/badge/skills/jairotorregrosa/jaiskills/goal-loop/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/jairotorregrosa/jaiskills/goal-loop"><img src="https://agentmods.dev/badge/skills/jairotorregrosa/jaiskills/goal-loop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00027 | $0.03253 |
| Opus 5.5 | $0.00011 | $0.01301 |
| Sonnet 5.5 | $0.00005 | $0.00651 |
| Haiku 4.5 | $0.00003 | $0.00325 |
Grade A, and why
goal-loop 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 1mo 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 — 172 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Goal Loop: Gradient Descent for Goals, Built by an Agent Factory
Objective and flags: $ARGUMENTS
Empty objective → ask for one. Parse the flags (table at the end) before Phase 0.
Formal contract (invariants, phase pre/postconditions, conformance checks): references/spec.md. Read it before Phase 0 of a first run and whenever a phase rule is unclear. Worked example, end to end: references/example-cv-os.md. Read it when the goal is frontend or judgment-heavy.
Treat the goal as a training problem. The working tree is the parameter, evidence commands are the loss function, a fresh implementer is the forward pass, a diagnoser produces textual gradients (the backward pass), momentum accumulates recurring error patterns, and an independent judge validates against held-out checks the implementer never sees. An agent factory generates all of these agents specialized to the goal, consulting an archive of past loops.
Deep-Learning Mapping
| Deep learning | This loop | Where it lives |
|---|---|---|
| Parameter θ | Working tree + artifacts | the repo |
| Forward pass | Fresh implementer attempt | agents/implementer.md |
| Training loss | Visible validation evidence | loop.md § Evidence (visible) |
| Held-out test set | Hidden compositional checks | heldout.md (judge-only) |
| Gradient ∂L/∂θ | Textual diagnosis: which behavior caused the failure, what to change | diagnoser output |
| Momentum | Recurring-pattern memory across epochs | loop.md § Momentum |
| Learning rate | Edit scope per epoch (shrinks on plateau) | contract field |
| Epoch | One full loop iteration | descent log |
| Early stopping | No loss improvement for patience epochs |
decide step |
| Train/test leak | Showing held-out checks to the implementer | forbidden |
Directory Layout
loops/
archive.md # cross-goal archive: agent designs + instruments that worked (stepping stones)
<slug>/
loop.md # goal contract + descent log — the implementer sees this
heldout.md # held-out evidence — NEVER included in any implementer prompt
agents/ # factory-generated, goal-specialized agent prompts
implementer.md
verifier.md
diagnoser.md
judge.md
tools/ # factory-generated instruments (probes, harnesses, generators)
What ships with it
7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago Changed · +7 lines · -191 tokens per session 3117464e2d32
- 1mo ago First seen · 165 lines · 218 tokens per session scan A 788cbc9fc805
goal-loop is a skill published in the GitHub repository JairoTorregrosa/jaiskills (10 stars, last pushed 3d ago), licensed MIT. It adds 27 tokens to every session and 3,253 once invoked, about $0.0001 per session on Opus 5.5. 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 skills, from other repositories
eliza
Skill "eliza" from aomi-labs/skills, covering elizaos, what you probably got wrong, installation, cli (recommended for new projects) and from source (contributor workflow).
swarmclaw
AI agent runtime and multi-agent orchestration platform. Teaches agents how to use SwarmClaw's 6 primitive tools, persistent memory, dreaming, delegation, connectors, credentials, and the skill system. Use when an agent is running on SwarmClaw and needs to understand the platform's capabilities.
update-harness
Update an existing Claude Code harness — check for outdated components, duplication with official features, missing patterns, and improvement opportunities, then apply approved changes. Use whenever: 'update harness', 'update my harness', 'is my harness up to date', 'harness outdated', 'optimize harness', 'check my…
create-harness
Generate a complete, project-specific Claude Code harness — agents, skills, hooks, CLAUDE.md rules — tailored to any codebase. Use whenever someone says 'create a harness', 'set up agents for this project', 'configure Claude Code', 'build an agent team', 'make this project AI-ready', 'harness this project', or any…
help
Get oriented with the solopreneur plugin — see your AI team, check project status, and get suggestions for what to do next. Use when you're getting started or need a refresher.
review
Review code, specifications, or designs for quality, bugs, security, and best practices. Use when the user wants feedback on recent work, a pull request, or any artifact.