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 instructions/zalom/plastic/agents-mdgit clone --depth 1 https://github.com/zalom/plasticWhat 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 | $0.02141 | $0.02141 |
| Opus 5 | $0.01071 | $0.01071 |
| Sonnet 5 | $0.00428 | $0.00428 |
| Haiku 4.5 | $0.00214 | $0.00214 |
Grade A, and why
plastic AGENTS.md 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 — 132 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Plastic
Intent-driven state management for AI coding sessions.
Stack
- Language: Ruby (scripts), JavaScript/Node.js (npm package, installer)
- Framework: npm package (CLI installer), flat personal skills
- Testing: Minitest. See the Testing section below for the correct full-suite command.
- Source: /Users/zlatko/apps/personal/plastic
- Remote: [email protected]:zalom/plastic.git
Defaults
- Release process: commit_and_push, github_release, npm_publish
- Version files: package.json, .claude-plugin/plugin.json, .claude-plugin/marketplace.json
- Tag format: v{{version}}
- All bash scripts must work under macOS /bin/bash 3.2 (no bash 4.x features)
- Bump all 3 version files on every fix/feature release
Searching Plastic with QMD
QMD is an optional but recommended local markdown search engine over the Plastic stores. Plastic works without it (ripgrep over the store files is the fallback). When QMD is present, search it before re-deriving an existing decision, spec, or outcome.
- Collections. Plastic indexes into the DEFAULT qmd index, one collection per store,
all
plastic-prefixed:plastic-globalfor~/.plastic/store, andplastic-<slug>for each project store (slugs come fromprojects.yml). - Scope a search. One project:
-c plastic-<slug>. All of Plastic: target theplastic-*collections. Pick the narrowest scope that answers the question. - Search before re-deriving. Look in the stores for prior decisions, specs, and outcomes before re-deriving them. The stores are the memory.
- Power tools are recommended when present. When QMD (for intents), Enola, or Serena
(for code navigation) is present, prefer them: the UserPromptSubmit hook appends one
combined recommendation line covering QMD and at most one code-navigation tool,
Enola-first when both code-navigation tools are present (one code-navigation slot,
Enola wins over Serena), and the per-skill
QMD-first steps run
qmd-sync searchbefore grep/Read, then open the authoritative intent file. Use the deterministicscripts/qmd-sync search "<terms>"helper, which scopes collections for you and is a clean no-op when QMD is absent. - Completion fires an async reindex. Intent delivery reindexes the delivering store's collection in the background (non-blocking), so the index stays fresh while "index mutation is lifecycle-only" stays true. Never reindex ad-hoc.
- Index mutation is lifecycle-only. The index changes only on Plastic lifecycle events (install registers all stores, project creation registers the new project store, intent delivery reindexes the delivering store's collection). Never reindex ad-hoc.
- Query craft lives in the qmd skill. Use the installed
qmdskill for structuredqmd query(intent/lex/vec/hyde) and BM25qmd search. Two notes: structured queries need ANSI-C$'...'quoting so\nbecomes a real newline, andqmd search(BM25) needs no model downloads, so it is the safe model-free fallback.
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 · 132 lines · 2,141 tokens per session scan A cebba0006993
plastic AGENTS.md is an instructions file published in the GitHub repository zalom/plastic (10 stars, last pushed 2d ago), licensed MIT. It adds 2,141 tokens to every session, about $0.0107 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 instructions, from other repositories
reka CLAUDE.md
Claude Code instructions for getreka/reka, covering claude.md, standing repo policy, build and run commands, rag api (main backend service) and mcp server.
project-os-for-codex AGENTS.md
Instructions for herry2059/project-os-for-codex, covering ai contributor rules, product truth, architecture and security, development workflow and agent integration contract.
AmbyKit CLAUDE.md
Instructions for ambystechcom/AmbyKit, covering claude code — notes for the ambykit repo and claude-specific.
sugar AGENTS.md
AGENTS.md instructions for roboticforce/sugar, covering sugar - agents.md, git branching rules (strict), always create branches from develop, after work is complete and then create pr targeting develop.
nextjs-skills CLAUDE.md
Instructions for mohamed-hossam1/nextjs-skills: All contribution rules — style, naming, file layout, verification — live in AGENTS.md. Read that first.
humanfile copilot-instructions.md
Instructions for zhangyu94/humanfile, covering .human file boundaries, rules and behavior.