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 skills add Yuchen20/Context-Crumb --skill contextcrumbgit clone --depth 1 https://github.com/Yuchen20/Context-CrumbWrote 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.
[](https://agentmods.dev/skills/yuchen20/context-crumb/contextcrumb)<a href="https://agentmods.dev/skills/yuchen20/context-crumb/contextcrumb"><img src="https://agentmods.dev/badge/skills/yuchen20/context-crumb/contextcrumb.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00043 | $0.01475 |
| Opus 5 | $0.00022 | $0.00737 |
| Sonnet 5 | $0.00009 | $0.00295 |
| Haiku 4.5 | $0.00004 | $0.00147 |
Grade A, and why
contextcrumb scanned grade A with 1 finding 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 8d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl https://huggingface.co/spaces/ymao20/contextcrumb-32m-demo/agents.md How it starts
The opening of the file, as written. The whole thing — 161 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ContextCrumb
Use ContextCrumb when large prose-heavy text would otherwise bloat the context window. It can also load supported source files safely in auto mode: executable code is preserved exactly, and only comments/docstrings are compressed. The compressed output is the working context, not a preview.
Do not read large candidate files with cat, Get-Content, type, or similar raw file dumps before compression. Inventory names first, then use contextcrumb load to read compressed context when the CLI is available.
Use For
- Documentation, Markdown, notes, transcripts, threads, logs, and research dumps
- Prose-heavy local text where compression is acceptable before LLM reading
- Supported source files when comment/docstring compression is useful and exact executable code must remain intact
- Folders of natural-language text files when you need sampled or batched context loading
Do Not Use For
- Unsupported source files, diffs, configs, schemas, commands, or structured data
- Legal text, policy text, contracts, or compliance material
- Any task that requires exact syntax, exact wording, exact formatting, or precise quotes
Quick Workflow
Decision tree:
- If reading a large local prose file, use local ContextCrumb.
- If reading a supported code file, use local ContextCrumb in default
automode or--content-mode code-comments; then read raw source before exact edits. - If local ContextCrumb is unavailable and the user provided pasted, non-sensitive text, use the hosted trial.
- If local ContextCrumb is unavailable and the input is a local or private file, ask the user to install local ContextCrumb.
- If exact wording, syntax, formatting, or every token matters, do not compress.
Inventory without reading file bodies:
rg --files <directory>
Check the CLI:
contextcrumb --help
If it is not on PATH, try the module form before installing:
python -m contextcrumb --help
If the project visibly uses an active virtual environment or tool-managed Python command, use that environment's normal command form rather than installing globally.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 8d ago First seen · 161 lines · 43 tokens per session scan A 18e7f3d75820
contextcrumb is a skill published in the GitHub repository Yuchen20/Context-Crumb (1 stars, last pushed 3mo ago), licensed MIT. It adds 43 tokens to every session and 1,475 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
find-skills
Use when automatically discover, evaluate, and activate community skills when local skills don't cover user needs. Includes credibility scoring and safety checks for complete OpenClaw self-sufficiency.
apple-notes
Interact with the Apple Notes app. CRUD operations for persistent storage of thoughts, data, and information across sessions.
cron
Use this skill only for scheduled or recurring tasks. Manage jobs with qwenpaw cron list/create/get/state/update/pause/resume/delete/run, and always pass --agent-id explicitly.
cron
A procedure for creating and managing tasks that run automatically at a future time or repeat on a schedule. It uses QwenPaw's scheduling commands and requires an agent identifier.
mailbox
A guide for performing email and mailbox tasks through the QwenPaw mailbox system. It covers supported personal email providers, credentials, configuration, and the required mailbox workflow.
channel_message
Use this skill to proactively send a one-way message to a user/session/channel, usually only when the user explicitly asks to send to a channel/session or when proactive notification is needed. First query sessions with qwenpaw chats list, then push with qwenpaw channels send.