strata:observe

strata:observe is a skill for Claude Code, Codex from gideondk/strata. It costs 94 tokens per session (587 once invoked), scanned A, original, MPL-2.0.

A skill for recording grounded, low-risk observations about a project in a temporary review area. An observation is a factual note tied to a real file or Git commit, rather than a decision or a definition.

In plain words
What is it for?
It is for recording facts such as which commit changed a policy or which folder owns a responsibility. It requires a source file or commit and does not record decisions, disputed questions, or domain definitions.
Why use it?
It preserves useful project facts without treating them as permanently approved knowledge or interrupting the user for every small note.

Skill for Claude CodeCodex

Installs and runs on its own, but its text points at files inside its plugin — anything it tells you to read at a ${CLAUDE_PLUGIN_ROOT} path is only there once the plugin is installed. Installing the plugin gets both.

Part of the strata plugin — 24 skills, 2 agents, 5 hooks, 1 MCP server shipped together

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/gideondk/strata/observe
Any agent
npx skills add gideondk/strata --skill observe
Clone the repo
git clone --depth 1 https://github.com/gideondk/strata

Made for: Claude Code, Codex.

Or install strata, the plugin that ships this one along with the rest of its 24 skills, 2 agents, 5 hooks, 1 MCP server.

Wrote 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.

agentmods badge for strata:observe

README.md
[![agentmods](https://agentmods.dev/badge/skills/gideondk/strata/observe.svg)](https://agentmods.dev/skills/gideondk/strata/observe)
Your own site
<a href="https://agentmods.dev/skills/gideondk/strata/observe"><img src="https://agentmods.dev/badge/skills/gideondk/strata/observe.svg" alt="Measured on agentmods" height="20"></a>
Per session 94 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 587 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 5.1 $0.00094 $0.00587
Opus 5 $0.00047 $0.00293
Sonnet 5 $0.00019 $0.00117
Haiku 4.5 $0.00009 $0.00059

Measured 5d ago against content hash 26fd84c68957, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

strata:observe 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 5d 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.

skills/observe/SKILL.md · 53 lines

What it actually says

strata:observe

The safe autonomous-write lane (see the autonomy-line ADR: gate on reversibility + grounding, never on confidence). An agent may write a grounded, reversible observation without asking — because markdown+git make it trivially revertible and it never ratifies anything.

When to use

  • A durable breadcrumb anchored to a real artifact: "commit abc changed the token-bucket policy", "billing/ owns invoice retries".
  • Worth keeping, but not worth interrupting the user for, and not a decision.

Never use for: decisions (/strata:decide), contested questions (/strata:propose), or domain definitions (/strata:domain). Those are human-ratified — that's the moat. This script structurally cannot write them.

How

Grounding is required — pass --source-file (project-relative, repeatable) and/or --commit. Ungrounded auto-writes are refused.

echo "Commit bumped the HTTP retry budget 3→5; see the client." | \
  "${CLAUDE_PLUGIN_ROOT}/bin/strata" observe \
  --topic "retry budget bumped" \
  --source-file "src/http/client.py" --commit "$(git rev-parse --short HEAD)"

The note lands in pr-context/<branch>/ with status: auto + source: git-derived. It announces itself and is quarantined from recall (it won't surface in /strata:find, the recall tool, or the primer) until a human reviews it.

The review lifecycle (keep or discard)

Auto-notes pile up in the "Awaiting your input" section of /strata:dashboard. For each:

  • Keep / promote — edit the note and remove the status: auto line (or change it to a normal status). It then enters recall like any other note.
  • Discard/strata:forget <path> (or just delete the file).

This is "machine proposes, human disposes": capture is automatic, but nothing unreviewed is ever served as canonical memory.

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. 5d ago First seen · 53 lines · 94 tokens per session scan A 26fd84c68957

Subscribe to this mod's changes

strata:observe is a skill published in the GitHub repository gideondk/strata (5 stars, last pushed 1mo ago), licensed MPL-2.0. It adds 94 tokens to every session and 587 once invoked, about $0.0005 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.

Related

Other skills, from other repositories

my-wiki

Manage local OKF-compatible Markdown My Wiki vaults with an AI agent. Use for capturing webpages, PDFs, Office documents, notes, images, folders, and ZIP bundles as References; maintaining Reference-to-Concept evidence links; searching or answering from a vault; checking or repairing vault health; switching among…

NimaChu/my-wiki · 78 tokens

kb

Set up, evolve, or operate a hraness/kb local-first Markdown knowledge base for coding-agent memory. Use when a user asks to design KB conventions or a recurring KB ritual; search or query a KB or Obsidian vault; load or recover repository context, plans, decisions, concepts, backlinks, semantic search, or Git…

hraness/kb · 155 tokens

mneme-search

Use when the user asks a factual question whose answer might live in the vault. Invokes mnemesearch. Production mnemesearch is FTS5 BM25. The experimental feature-hashed lexical-vector backend is not wired into MCP. KG enrichment is gated to summarize or timeline when full-profile graph state is active.

OnourImpram/mneme · 69 tokens

mneme-prime

Use when the user starts a new task that may have prior vault context worth surfacing. Invokes the mnemeprime MCP tool to build a token-budgeted preamble of recent sessions and topic matches.

OnourImpram/mneme · 46 tokens

memory-forest

Build, validate, index, and query a private local Memory Forest with provenance-aware layers, pointer-first search, and validated root-first retrieval. Use when creating a new forest, checking an existing forest's structure, building its local SQLite index, or retrieving bounded memory trails without exposing full…

hyungchulc/memory-forest · 65 tokens

brain

Orient in the user's big-brain second brain and work brain-aware for the rest of the session. Use when the user types /brain, or asks "what am I working on", "what's on my plate", "catch me up", "what's overdue", or at the start of a session about ongoing work. Loads the vault overview, reports it in one screen, and…

jzhao234/big-brain · 87 tokens