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.
git clone --depth 1 https://github.com/frankxai/Starlight-Intelligence-SystemWrote 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/commands/frankxai/starlight-intelligence-system/training-measure-transfer)<a href="https://agentmods.dev/commands/frankxai/starlight-intelligence-system/training-measure-transfer"><img src="https://agentmods.dev/badge/commands/frankxai/starlight-intelligence-system/training-measure-transfer/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/commands/frankxai/starlight-intelligence-system/training-measure-transfer"><img src="https://agentmods.dev/badge/commands/frankxai/starlight-intelligence-system/training-measure-transfer.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 | $0.00061 | $0.02381 |
| Opus 5 | $0.00030 | $0.01190 |
| Sonnet 5 | $0.00012 | $0.00476 |
| Haiku 4.5 | $0.00006 | $0.00238 |
Grade A, and why
training-measure-transfer 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 6d 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 — 194 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/training-measure-transfer
Load agents/starlight-training.md, skills/people-intelligence/learning-architecture.md, the Curriculum Design, and the Program Design. Produce the Transfer Measurement Plan — the load-bearing artifact that determines whether the program can prove transfer.
Input
$ARGUMENTS
Flags
--post-window <30|60|90|180>— required. The L3 measurement window closes at this point post-program-end. Default 90.--baseline-method <observation|scenario-test|peer-rating|metric-snapshot|hybrid>— required. The pre-program baseline mechanism. Without baseline, transfer cannot be claimed.
Process
-
Verify baseline exists. If no baseline measurement was captured before program delivery, halt. The program cannot prove transfer. Options: (a) defer launch and capture baseline now; (b) launch with explicit caveat that no transfer claim will be made; (c) refuse.
-
Refuse L1-only. If the requester's measurement plan is "we'll send a survey at the end," refuse the framing. Smile-sheets are noise; they are not evaluation. Insist on L3 minimum.
-
L3 design. Behavior observation protocol:
- Rubric: the L3 rubric from the Curriculum Design — observable behavioral indicators per L3 sub-behavior.
- Sampling: who observes (manager + peer + sometimes self with behavioral examples), how often (30/60/90 minimum), how many instances per learner per checkpoint.
- Triangulation: at least two observation sources per learner per checkpoint. Manager-only is biased; manager + peer is stronger; manager + peer + self-with-examples is strongest.
- Inter-rater reliability check: for high-stakes programs, calibrate observers on the rubric before the first measurement window.
-
L4 design. Business-metric tracking:
- Metric source: where the metric lives (engagement survey, retention dashboard, NPS pipeline, quality KPI system, etc.).
- Baseline value: captured at T-0.
- Target value: projected at T+post-window.
- Confounders: explicit list — other initiatives, market factors, seasonality, org changes — that could move the metric independently of training. Acknowledge; do not over-claim causation.
- Comparison group (if feasible): ideal but rarely available; if available, structure pre/post + comparison-group analysis.
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.
- 6d ago First seen · 194 lines · 61 tokens per session scan A 1b60f740241d
training-measure-transfer is a command published in the GitHub repository frankxai/Starlight-Intelligence-System (8 stars, last pushed today), licensed MIT. It adds 61 tokens to every session and 2,381 once invoked, about $0.0003 per session on Opus 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 2026-09-03.
Other commands, from other repositories
memories
View and manage learned memories.
mpm-session-resume
Load context from paused session.
forget
Delete specific memories.
learn
Add new learning to memory.
gbu-retro
Post-session retrospective — harvest this session's lessons into durable doctrine.
agent-list
List all available agents with their capabilities (MPM-provided — may be shadowed by a user /agent-list command).