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/avinash-1707/strata/agents-mdgit clone --depth 1 https://github.com/avinash-1707/StrataWhat 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 | $0.01208 | $0.01208 |
| Opus 5 | $0.00604 | $0.00604 |
| Sonnet 5 | $0.00242 | $0.00242 |
| Haiku 4.5 | $0.00121 | $0.00121 |
Grade A, and why
Strata 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 yesterday.
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 — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Strata is a local-first MCP server giving AI agents compressed, searchable memory. Node 20 + TypeScript, Postgres + pgvector, Redis, Ollama. Single user, single host, no external API calls.
Which document wins
docs/ is authoritative. context.md is the original spec and is known to
contain four critical defects that docs/ corrects — including a write path
that can lose memories and a forget that doesn't forget. Where the two differ,
follow docs/ and the referenced DD-0NN. Never implement context.md
verbatim.
Read only what the task needs
Do not load any document wholesale. Load the narrowest set that covers the task:
| Task | Read |
|---|---|
Writing or changing any code in src/ |
docs/coding-standards.md |
| Adding/changing a tool, module, or data flow | docs/architecture.md + docs/coding-standards.md |
| "What do I build next?" | docs/build-plan.md (current phase only) + docs/progress-tracker.md (Current state block only) |
| Database schema, migrations, indexes | docs/architecture.md § Data model |
| Search, ranking, or fusion work | docs/architecture.md § Retrieval pipeline |
| Prompts, LLM parsing, Ollama calls | docs/architecture.md § Model layer |
| Docker, Compose, env vars, deployment | docs/architecture.md § Runtime topology |
| "Why is it done this way?" / proposing a change to a settled choice | docs/design-decisions.md |
| Onboarding, scope questions, "should we build X?" | docs/project-overview.md |
| Reviewing or finishing someone else's work | docs/progress-tracker.md (Current state) + the phase's entry in docs/build-plan.md |
Rules for reading:
- Read
docs/progress-tracker.md's Current state block only. The Log below it is history — skip it unless you need to know why something changed. - Read only the current phase of
docs/build-plan.md, not all phases. - Consult
context.mdonly when a doc contradicts itself or omits a spec detail.
Always applies
These hold for every code change. Details in docs/coding-standards.md.
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.
- yesterday First seen · 91 lines · 1,208 tokens per session scan A f14ee807feef
Strata AGENTS.md is an instructions file published in the GitHub repository avinash-1707/Strata (0 stars, last pushed 25d ago), licensed MIT. It adds 1,208 tokens to every session, about $0.0060 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
libre-webui AGENTS.md
Instructions for libre-webui/libre-webui, covering agents.md, project overview, project structure, setup and development and testing and validation.
paperless-brain CLAUDE.md
Instructions for vailsen/paperless-brain, covering claude.md, what this is, running, local dev and app runs at http://0.0.0.0:8080.
opencrawling CLAUDE.md
Instructions for opencrawling/opencrawling, covering claude.md, project overview, build commands, full build (all maven modules) and run the embedding microservice (separate process).
vector-mcp AGENTS.md
Instructions for Knuckles-Team/vector-mcp, covering agents.md, scope, current architecture, non-negotiable design rules and source layout.
PBP CLAUDE.md
Instructions for MadGapun/PBP, covering pbp — claude-code-memory, stand 2026-08-25 (v1.7.23 stable) — ehrliche zahlen, stand 2026-07-16 (v1.8.0-beta.6, prerelease) — hotfix #760, stand 2026-07-14 (v1.7.7) — scoring-fairness & praxis-funde and stand 2026-07-03 (v1.7.6) — alltags-fuehrung.
artel CLAUDE.md
Instructions for NicolasPrimeau/artel, covering artel, what it is, stack, layout and core primitives.