core

A framework for organising research and audits across several AI models and work sessions. PRISM is the name of the framework and its files and procedures.

In plain words
What is it for?
Use it when starting or continuing a PRISM audit, loading its central operating rules, preparing prompts, or managing research phases and continuity.
Why use it?
It provides a shared structure for complex investigations that span multiple sessions, models, and source files. The description does not specify a particular research topic or output.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/ronkupper/prism/core
Any agent
npx skills add Ronkupper/PRISM --skill core
Clone the repo
git clone --depth 1 https://github.com/Ronkupper/PRISM

Made for: Claude Code, Codex.

Per session 269 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,084 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00269 $0.02084
Opus 5 $0.00134 $0.01042
Sonnet 5 $0.00054 $0.00417
Haiku 4.5 $0.00027 $0.00208

Measured 2d ago against content hash 595ef77b6028, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

core 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.

plugins/prism/skills/core/SKILL.md · 141 lines

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.

PRISM framework loader

This Skill loads PRISM — the canonical operating framework for structured, multi-session, multi-vendor LLM-orchestrated audit and research. The framework is decomposed into a lean always-loaded core plus on-demand bundles; load the core first, fetch a bundle only when the work in front of you needs it.

When triggered

  1. Read PRISM_core.md in full before acting — it is the operating core (operating model and architecture, the Execution Envelope / Self-check / Output triple contract, the prompt-package engine, the context-pressure framework including the M5 bands and the in-core recovery floor, the §{section.lens-schema-what-orchestration-consumes} lens schema, Monitors M1–M12, Standing Principles, filename conventions, the atomic-prompt template, the operator-hint catalog, the engagement-closure gate, and the phase → bundle manifest). Phase-scoped mechanics (Setup, currency, continuity, corpus-access) are on-demand bundles — see step 5.
  2. Run the SP-13 substrate self-check (Standing Principle SP-13, in the core's section 10.1.3): declare your model identity and confirm you are Claude, Opus-class / flagship tier (a capability floor — version-agnostic, latest by default). Halt and ask the operator on mismatch or cannot-determine.
  3. Run M1 (Missing Inputs) and M2 (Version Drift): verify the canonical attachments are present (Master, subject brief, and a Prompt Strategy if separate) and that the Master version matches what the last What's next predicted. Halt at HIGH severity on a missing required artifact or a version mismatch.
  4. Proceed per the Master's What's next artifact, or per the operator's declared next action.
  5. Load the phase bundle for the role you occupy. The core is lean; phase-, seam-, and condition-scoped mechanics are fetched on demand per the core's phase → bundle manifest (section 3.7.6). Run the visible self-check and fetch:
    • Setup / P0 (no Master yet, or the operator initiates an engagement) → reference/setup.md (probes, three-layer readiness, Setup-artifact procedures).
    • Update session / the session-open currency checkreference/currency.md.
    • Resume with a missing or abnormal handoff, or M5 band 🟠/🔴reference/continuity.md (the in-core recovery floor stays resident; the full migration / recovery procedure is in the bundle).
    • A pass classified corpus-accessreference/corpus-access.md. Each gutted area carries a fail-loud in-core stub — operating without the required bundle is detectable, not a silent gap.

Read the full file on GitHub · 141 lines

Changes

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.

  1. 2d ago First seen · 141 lines · 269 tokens per session scan A 595ef77b6028

Subscribe to this mod's changes

core is a skill published in the GitHub repository Ronkupper/PRISM (5 stars, last pushed 2mo ago), licensed MIT. It adds 269 tokens to every session and 2,084 once invoked, about $0.0013 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.