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 0spoon/seamless --skill seam-researchgit clone --depth 1 https://github.com/0spoon/seamlessWrote 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/0spoon/seamless/seam-research)<a href="https://agentmods.dev/skills/0spoon/seamless/seam-research"><img src="https://agentmods.dev/badge/skills/0spoon/seamless/seam-research/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/0spoon/seamless/seam-research"><img src="https://agentmods.dev/badge/skills/0spoon/seamless/seam-research.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 | $0.00065 | $0.00836 |
| Opus 5.5 | $0.00026 | $0.00334 |
| Sonnet 5.5 | $0.00013 | $0.00167 |
| Haiku 4.5 | $0.00006 | $0.00084 |
Grade A, and why
seam-research 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 — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Run a Seamless research lab
Use Seamless's lab_open, trial_record, and trial_query MCP tools to keep an
immutable, structured lab notebook shared with every agent investigating the
same problem.
Codex invokes this skill as $seam-research <lab-name> <problem>; Claude Code
invokes it as /seam-research <lab-name> <problem>. Use the arguments from the
user's invocation. Claude Code may also expand them here: $ARGUMENTS.
Parse the first word as the lab name and the remainder as the goal. With only a lab name, resume that investigation. If the skill activated implicitly with no arguments, derive a short stable name, tell the user, and continue.
1. Open or resume the lab
Call lab_open with the name and goal. It binds the lab to this MCP connection
and returns recent trials. If trials exist, call trial_query for the full
history and summarize:
- trial count and observed outcome groups;
- hypotheses the evidence eliminates;
- durable decision memories mentioning the lab, found with
recall; - the last trial and the most useful next unknown.
2. Run the evidence loop
Before each trial:
- Call
trial_query; filter by outcome or metrics when the history is long. - Identify the variable being isolated and what prior trial makes it useful.
- State the predicted outcome before running the experiment. Include a confidence level, quantitative expectation when possible, and the result that would change the model.
After observing the result, call trial_record exactly once. Trial records are
immutable. Supply:
title: short and specific;changes: exact files, settings, commands, or hardware changes;expected: the prediction stated before the run, unchanged;actual: observed behavior, measurements, and relevant log lines;outcome:pass,fail,partial, orinconclusive;metrics: queryable numeric or categorical values as a structured object.
Record an abandoned trial as inconclusive with the blocker so another agent
does not repeat it blindly. Compare expected and actual immediately; a mismatch
is evidence, not a footnote.
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 First seen · 82 lines · 65 tokens per session scan A 18a3a025ea86
seam-research is a skill published in the GitHub repository 0spoon/seamless (5 stars, last pushed 25d ago), licensed MIT. It adds 65 tokens to every session and 836 once invoked, about $0.0003 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
agent-state-bar
A design guide for an agent status bar: a short, automatically updated summary of progress, remaining tasks, tool counts, errors, and system state placed near the latest model input.
kv-cache-design
A guide to arranging an AI agent's conversation so previously processed text can be reused. A KV cache stores intermediate results for earlier tokens, while a prompt cache reuses a stable beginning across requests.
reward-design
A guide to defining the scores that an AI agent receives during trial-and-error training. It covers checks based on rules, human preferences, or another model, and distinguishes rewarding the final result from rewarding the steps taken.
agent-evaluation
A guide to measuring whether an AI agent works reliably on the tasks that matter. It covers repeatable test environments, program checks, human-like scoring rules, failure analysis, and experiments that separate model problems from system problems.
post-training-strategy
A guide to choosing and planning ways to improve an AI model after its initial training. It explains when to add knowledge, teach a response format, use trial-and-error training, or use a smaller model to learn from a larger one.
ai-style
A Chinese-language writing guide for producing or revising reader-facing text such as product announcements, public-account articles, emails, and README files. It focuses on reducing patterns that make writing sound machine-generated.