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 skills/craftos-dev/craftbot/memory-processornpx skills add CraftOS-dev/CraftBot --skill memory-processorgit clone --depth 1 https://github.com/CraftOS-dev/CraftBotWrote 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/craftos-dev/craftbot/memory-processor)<a href="https://agentmods.dev/skills/craftos-dev/craftbot/memory-processor"><img src="https://agentmods.dev/badge/skills/craftos-dev/craftbot/memory-processor.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 | $0.00016 | $0.02620 |
| Opus 5 | $0.00008 | $0.01310 |
| Sonnet 5 | $0.00003 | $0.00524 |
| Haiku 4.5 | $0.00002 | $0.00262 |
Grade A, and why
memory-processor 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 — 247 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Memory Processor
The only way for agent to save event into long-term memory.
Files
agent_file_system/EVENT_UNPROCESSED.md- Source (read & clear batches)agent_file_system/MEMORY.md- Destination (append distilled memories)
Todo Tracking (REQUIRED)
Use update_todos to track progress. Create todos at start, update as you go.
Initial todos (create after reading first batch):
1. [pending] Process and loop for each batch (25~50 lines).
2. [pending] Validate and cleanup
3. [pending] Complete task
After each batch: Mark current as completed, add next batch if more events exist.
Batch Processing Workflow
Process 50 lines at a time to avoid memory issues.
Steps:
- Read first batch:
read_fileEVENT_UNPROCESSED.md, offset=11, limit=50 - Create todos: Use
update_todosto create initial todo list - Loop for each batch:
- Distill batch: Apply rules below, extract IMPORTANT memories only
- Append memories:
stream_editMEMORY.md (append only) - Remove batch:
stream_editEVENT_UNPROCESSED.md (delete lines 12-61) - Update todos: Mark batch completed, add next batch if more events
- Validation (mark todo in_progress):
- Validate no more unprocessed events in EVENT_UNPROCESSED.md
- Validate no duplicated memory in MEMORY.md
- End task:
end_turnwhen validation passes
Rules
- Silent background task. NEVER use send_message or interact with user.
- Immediately discard these event types:
[reasoning],[action_start],[gui_action],[screen_description][agent message]- agent responses are NEVER saved- Greetings, small talk, acknowledgments ("hi", "thanks", "ok")
- Screen descriptions ("The current screen displays...")
- Truncated text ending in
...
[user message]lines are NEVER discarded by type. They are the PRIMARY source of memories: apply the distillation rules to the CONTENT of every[user message]line. A user message that contains a preference, fact, contact, decision, or dated event MUST produce a distilled memory even when it is phrased inside an ordinary request (e.g. "I'm allergic to peanuts, find me a lunch spot" → save the allergy, discard the request).
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 · 247 lines · 16 tokens per session scan A f24d109a5e2b
memory-processor is a skill published in the GitHub repository CraftOS-dev/CraftBot (378 stars, last pushed 2d ago), licensed MIT. It adds 16 tokens to every session and 2,620 once invoked, about $0.0001 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 2026-09-03.
Other skills, from other repositories
context-pack-builder
當需要把長對話、專案歷史、研究資料或零散筆記濃縮成可交接給 Codex、其他 AI Agent 或人類接手的 Context Pack 時使用。.
ast-skeletonizer
How to extract high-density structural code skeletons (types, interfaces, class definitions, and function signatures) while stripping implementations to reduce context token consumption by 85%.
productivity-memory
Two-tier memory system that makes the AI a true workplace collaborator. Decodes shorthand, acronyms, nicknames, and internal language so the AI understands requests like a colleague would. Uses CONTEXT.md for working memory and a memory/ directory for the full knowledge base.
productivity-update
Sync tasks and refresh memory from your current activity.
shellm
Reference for the shellm system — recursive LLM shell, identity management, memory, skills, trajectory, and all CLI tools. Use when working on shellm itself, debugging agent behavior, or understanding how the pieces fit together.
resume
Load context from permanent memory and recent session logs. Tiered loading — CLAUDE.md full, session logs frontmatter-only by default.