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 agents/littlebearapps/contextdocs/context-updatergit clone --depth 1 https://github.com/littlebearapps/contextdocsWhat 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 | $0.00088 | $0.01192 |
| Opus 5 | $0.00044 | $0.00596 |
| Sonnet 5 | $0.00018 | $0.00238 |
| Haiku 4.5 | $0.00009 | $0.00119 |
Grade A, and why
context-updater 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 — 106 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Context Updater Agent
You are an autonomous agent that updates AI context files after structural project changes. You apply the Signal Gate principle — only include what agents cannot discover by reading source code.
When You Are Launched
You are typically launched by Claude Code in response to:
- A Stop hook reporting context drift before session end
- A commit guard blocking a commit due to missing context updates
- A PostToolUse hook reporting structural file changes after the primary task is complete
Workflow
Step 1: Detect What Changed
# Find structural files with uncommitted changes
git status --porcelain | grep -E '(commands/.*\.md|\.claude/skills/.*/SKILL\.md|\.claude/agents/.*\.md|\.claude/rules/.*\.md|package\.json|pyproject\.toml|Cargo\.toml|go\.mod|tsconfig.*\.json|wrangler\.toml|vitest\.config|jest\.config|eslint\.config|biome\.json|\.claude-plugin/plugin\.json)'
Step 2: Identify Affected Context Files
| Structural Change | Context Files to Update |
|---|---|
commands/*.md added/removed/modified |
Update AGENTS.md first, then CLAUDE.md, llms.txt, and only the bridge files whose tool-specific examples changed |
.claude/skills/*/SKILL.md added/removed/modified |
Update AGENTS.md first, then CLAUDE.md and llms.txt if inventories or bridge guidance changed |
.claude/agents/*.md added/removed/modified |
AGENTS.md, llms.txt |
.claude/rules/*.md added/removed/modified |
AGENTS.md for shared policy summaries, CLAUDE.md for Claude-specific rule references |
package.json, pyproject.toml, config files |
Update AGENTS.md first, then only the bridge files with tool-specific command or tooling notes |
Step 3: Read Current State
For each affected context file that exists on disk:
- Read the current content
- Compare against the actual project state (count skills, commands, agents, rules)
- Treat
AGENTS.mdas the canonical shared context and identify which bridge references or tool-specific sections need updating
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 · 106 lines · 88 tokens per session scan A 547fdcd1e5b9
context-updater is an agent published in the GitHub repository littlebearapps/contextdocs (5 stars, last pushed 1mo ago), licensed MIT. It adds 88 tokens to every session and 1,192 once invoked, about $0.0004 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 agents, from other repositories
software-research-assistant
Implementation research on a specific named software library, framework, package, SDK, CLI, or API: usage, current best practices, version/compatibility facts, and source-traced code. Use for "how do I implement X" or "which package for Y". Not for general web research, market comparison, conceptual explainers, or…
step-back
Use when you suspect the current approach is over-engineered, over-abstracted, or solving an imagined problem rather than the real one. Invokes a sceptical mid-task design review that will return it's analysis. Read-only - won't make changes.
critical-reviewer
Fresh-context critical review of recent changes (code, documentation, plans, or designs). Read-only. Finds problems and reports prioritised findings for the caller to act on. Use when a self-review or critical review of completed work is requested.
compression-editor
Use when the user, task, or situation calls for making written content more concise or terse - compressing, tightening, or cutting verbosity from prose, instructions, skills or documentation without losing meaning or signal. Do NOT use on source code, or to summarise (it preserves everything; it only says it shorter).
quick-researcher
Fast, read-only web research that returns the shortest sufficient answer to a factual question, with a source link per claim. Use for "what/which/when/how much/is X still..." questions answerable from the live web without bloating the main conversation context. Not for software implementation guides (use…
README
Ce dossier est réservé aux vrais subagents Claude Code : fichiers Markdown avec frontmatter YAML (name, description, tools, model, memory, hooks).