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 skills/mrbinnacle/azimuth/reference-authoringnpx skills add MrBinnacle/azimuth --skill reference-authoringgit clone --depth 1 https://github.com/MrBinnacle/azimuthWhat 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 | $0.00116 | $0.01478 |
| Opus 5 | $0.00058 | $0.00739 |
| Sonnet 5 | $0.00023 | $0.00296 |
| Haiku 4.5 | $0.00012 | $0.00148 |
Grade A, and why
reference-authoring 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 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.
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 — 141 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Reference File Authoring Discipline
Problem
Each new domain addition to AZIMUTH risks leaving Module 7 without vocabulary for the new domain — or creating redundant files when extension was the right operation. The symptoms: Module 7 produces generic output for a domain, or the analyst must load two files to complete one domain analysis.
EXTEND vs. CREATE Decision
EXTEND an existing file when:
- Adding ≤ 3 new patterns (≤ 30% increase on a 10-pattern file)
- The majority of existing patterns already apply to the new domain
- The file's own load condition already covers the new domain
CREATE a new file when:
- Adding 6+ patterns that are distinct from the existing file
- The new domain requires different load conditions than the host file
- Loading both the existing file and new content would require two separate loads anyway
Example applied: Migration patterns (3 patterns, 9 of 10 existing patterns apply to
migrations) → EXTEND software-failure-patterns.md. Hiring patterns (6 patterns,
zero overlap with software failure modes) → CREATE hiring-failure-patterns.md.
At 4–5 patterns, use judgment: if the domain is structurally distinct and all patterns are domain-specific (not shared with the host file), CREATE. If the domain shares most of its mechanisms with the host file, EXTEND.
Module 7 Vocabulary Header
Every new reference file and template must answer this question in its header:
"What vocabulary and diagnostic anchors does this file supply to Module 7?"
Without this, each new domain leaves Module 7 producing generic output — the analyst gets no domain-specific signal even when the file is loaded.
Format (include near the top of the file, after the load condition):
**Module 7 vocabulary this file supplies:** [list of specific terms, base rates,
and diagnostic anchors — e.g., "coachability as a distinct failure mode (26% of
failures); interpersonal vs. technical failure rate inversion (81% vs. 11%)"]
Why it matters: The gap scanner found that the hiring and startup domains both hit the same failure — Module 7 expected template-supplied domain vocabulary that wasn't there. The Module 7 header makes this provision explicit and checkable.
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 · 141 lines · 116 tokens per session scan A 2fe29fba849f
reference-authoring is a skill published in the GitHub repository MrBinnacle/azimuth (8 stars, last pushed 2mo ago), licensed MIT. It adds 116 tokens to every session and 1,478 once invoked, about $0.0006 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 skills, from other repositories
premortem
Run a premortem on any plan, launch, product, hire, strategy, or decision. Imagines it failed 6 months from now, works backward to find every reason why, then produces a revised plan. Triggers include "premortem this", "premortem my", "what could kill this", "stress test this plan", "find the blind spots", "poke holes…
pant-executive-decision-review
Pressure-test a product recommendation for an executive decision and rehearse the review. Use for approval, sponsorship, investment, priority, capacity, or commitment requests; not general roadmap or backlog review.
pant-business-case-review
Pressure-test a product investment business case and rehearse its review. Use for benefits, costs, alternatives, economics, uncertainty, adoption, ownership, and decision conditions; not a whole business model.
pant-meeting-rehearsal
Rehearse a consequential product conversation when no specialist PANT scenario fits. Use for realistic stakeholder dialogue, pressure changes, coaching, persuasion, and debrief; not artifact-first review.
pant-roadmap-review
Pressure-test a product roadmap and rehearse roadmap review conversations. Use for strategic coherence, outcomes, sequencing, capacity, tradeoffs, dependencies, adoption, and evidence; not sprint or backlog review.
pant-strategy-review
Pressure-test product strategy and rehearse strategy reviews. Use for diagnosis, choices, coherence, differentiation, capabilities, tradeoffs, market assumptions, and measures; not roadmap sequencing or execution plans.