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/hire-calibrate)<a href="https://agentmods.dev/commands/frankxai/starlight-intelligence-system/hire-calibrate"><img src="https://agentmods.dev/badge/commands/frankxai/starlight-intelligence-system/hire-calibrate/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/hire-calibrate"><img src="https://agentmods.dev/badge/commands/frankxai/starlight-intelligence-system/hire-calibrate.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 | $0.00086 | $0.03236 |
| Opus 5 | $0.00043 | $0.01618 |
| Sonnet 5 | $0.00017 | $0.00647 |
| Haiku 4.5 | $0.00009 | $0.00324 |
Grade A, and why
hire-calibrate 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 11d 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 — 217 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/hire-calibrate
This is part of the People Intelligence reference vertical. Composes with Genius Profile + Vision/Brand for company-as-candidate framing.
Load SIP.md, VOICES.md, agents/starlight-hiring.md, skills/people-intelligence/structured-hiring.md, the existing ICP (people-intelligence/hiring/icp-<role-slug>-*.md) and interview architecture (people-intelligence/hiring/interview-<role-slug>-*.md). Produce a Calibration Session Script — a facilitator-ready 60-minute agenda. Hand off to running the actual loop.
Disclaimer (non-waivable)
Hiring decisions touch employment law and protected-class considerations. This is system architecture, not legal advice. The calibration session itself does not surface legal questions; question stems must already have been reviewed by counsel before this session runs.
This command produces the script. The facilitator runs the live session. The team enters the loop calibrated, not running on instinct.
Input
$ARGUMENTS
Flags
--rater-count <3|4|5|6+>— number of raters in the calibration session. Minimum 3 (per Project Oxygen, cross-rater alignment requires ≥3 reference points). 4-5 is the sweet spot. 6+ creates discussion drag.
Process
-
Disclaim. Open with the non-waivable disclaimer.
-
Locate. Confirm role-slug. Read the ICP and interview architecture. If either is missing, halt and route to upstream command.
-
Identify anchor candidates. Two candidates the rater team has previously interviewed for similar roles — one retrospective hire-yes (worked out), one retrospective hire-no (or worked out poorly). The calibration session scores these against the new rubric. Their actual outcomes are the calibration anchor.
-
Build the 60-minute agenda.
- 0:00 - 0:05 (5 min) — Frame the session. Why calibration matters. Project Oxygen finding: cross-rater alignment beats rater quality. Decision: this team will not run an uncalibrated loop.
- 0:05 - 0:20 (15 min) — Rubric walk-through. Each rater paraphrases what 1, 3, and 5 mean for each dimension. Surface mismatched paraphrases. Agree on language.
- 0:20 - 0:40 (20 min) — Anchor candidate scoring. Each rater independently scores the two anchor candidates against the new rubric (8 min independent). Then surface scores in plenary (12 min). Where did raters disagree by ≥2 points on any dimension? Discuss those specifically.
- 0:40 - 0:55 (15 min) — Hire-bar agreement. What does a 3-on-this-dimension look like in our actual team? What does a 5-on-this-dimension look like? Agree on examples for each load-bearing dimension.
- 0:55 - 1:00 (5 min) — Question stem commitment. Each rater commits to using the agreed first-question stems verbatim. This kills divergent question framing across raters.
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.
- 11d ago First seen · 217 lines · 86 tokens per session scan A 83218fdbcd28
hire-calibrate is a command published in the GitHub repository frankxai/Starlight-Intelligence-System (8 stars, last pushed today), licensed MIT. It adds 86 tokens to every session and 3,236 once invoked, about $0.0004 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 2026-08-31.
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.
consolidate
Write a compact checkpoint summary of the current frontier.