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 Chemaclass/agnostic-ai --skill memory-curatorgit clone --depth 1 https://github.com/Chemaclass/agnostic-aiWrote 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/chemaclass/agnostic-ai/memory-curator)<a href="https://agentmods.dev/skills/chemaclass/agnostic-ai/memory-curator"><img src="https://agentmods.dev/badge/skills/chemaclass/agnostic-ai/memory-curator/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/chemaclass/agnostic-ai/memory-curator"><img src="https://agentmods.dev/badge/skills/chemaclass/agnostic-ai/memory-curator.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00031 | $0.00780 |
| Opus 5.5 | $0.00012 | $0.00312 |
| Sonnet 5.5 | $0.00006 | $0.00156 |
| Haiku 4.5 | $0.00003 | $0.00078 |
Grade A, and why
memory-curator 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 6d 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 — 45 lines — stays where its author put it; the contents beside it link to each section on GitHub.
memory-curator
Curates the auto memory store of the tool you are running in. The tool prunes its own store, during its own session, with the user watching.
Store
::target claude
Auto memory usually lives in ~/.claude/projects/<project>/memory/, but the location is configurable: autoMemoryDirectory in any settings scope moves it, CLAUDE_CONFIG_DIR moves the whole config root, and CLAUDE_CODE_PROJECT_DIR_NAME changes the project segment. Never assume the default path. Run /memory, pick the auto memory folder, and curate whatever store it opens. With the memory built-in on, sync sets autoMemoryDirectory to .agnostic-ai/local/memory/, the personal store every tool shares, so Claude's own saves land there and other tools read them. MEMORY.md is the index and every other file holds one topic.
Each session loads the first 200 lines or 25KB of MEMORY.md, whichever comes first. Everything past that is dropped on load.
::end
::target qoder
Auto memory lives in ~/.qoder/projects/<project>/memory/, with a user-level store in ~/.qoder/memory/. Run /memory for the overview and /memory manage to view, edit, or delete topic files. Ask which of the two stores to curate before reading anything, and curate one per run. MEMORY.md is the index and every other file holds one topic.
Qoder CLI reads the first 200 lines or about 25KB of each active MEMORY.md. Everything past that is dropped on load.
::end
Steps
- Find the store first. Open it through the tool's own
/memorycommand rather than a hard-coded path, and confirmMEMORY.mdexists there. If auto memory is turned off, say so and stop: there is nothing to curate until it is enabled. - Read
MEMORY.mdand list the topic files beside it. - Measure the index in lines and bytes against the load limit above, and report the space left.
- Group entries that state the same fact in different words. Those are merge candidates.
- Flag stale entries: a decision later work reversed, a date that has passed, a path or command the repository no longer has, a preference the user has since contradicted. Check each claim against the repository before calling it stale.
- Flag entries that belong somewhere else: anything you can work out from the code, and anything the instructions files already say. The session already carries those.
- Report one table: file, entry, verdict (keep, merge, delete, move), and one line of evidence.
- Propose the edits. Show the replacement line for a merge, and name the file and entry for a deletion.
- Stop there. Apply nothing until the user confirms which proposals to take.
- Apply only the confirmed items, then re-measure the index and report the new size.
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.
- 6d ago Changed 9b2545ab4d18
- 8d ago Changed b7721ad6e1e7
- 27d ago First seen · 45 lines · 31 tokens per session scan A 3ae0e9680d38
memory-curator is a skill published in the GitHub repository Chemaclass/agnostic-ai (25 stars, last pushed today), licensed MIT. It adds 31 tokens to every session and 780 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-09-19.
Other skills, from other repositories
session-end
Close a session — summarize changes, suggest commits, write a journal entry (episodic trace), write fresh NEXT.md (handoff). Chains to dream for substantive sessions.
dream
Sleep-consolidation pass — replay journal entries to distill them into garden/essential/USER.md, garden/essential/SOUL.md, garden/ .md, and new procedures in forge/ (all T1 autonomous within capacity). Propose T2 graduations to 3-control/foundation/SOUL.md, 3-control/foundation/PRINCIPLE.md, and 3-control/rule/ for…
session-start
Load read order into agent context at every session begin. Invoke detect-runtime in drift-check mode and propose re-ratify on runtime mismatch with profile.md. Consume + clear NEXT.md (the working buffer). Brief Owner on state. Not the one-time bootstrap (use init for that).
tldr-compress
Compress natural language memory files (CLAUDE.md, todos, preferences) into TLDR format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /tldr-compress FILEPATH or "compress…
agent-v3-memory-specialist
Agent skill for v3-memory-specialist - invoke with $agent-v3-memory-specialist.
memory-retain
Save what was learned to Hindsight long-term memory. Use when a task finishes, when the user states a fact, decision or preference, and when you discover something a later run or another Bot would need.