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 praveen-ilangovan/pensieve --skill claudegit clone --depth 1 https://github.com/praveen-ilangovan/pensieveWrote 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/praveen-ilangovan/pensieve/claude)<a href="https://agentmods.dev/skills/praveen-ilangovan/pensieve/claude"><img src="https://agentmods.dev/badge/skills/praveen-ilangovan/pensieve/claude/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/praveen-ilangovan/pensieve/claude"><img src="https://agentmods.dev/badge/skills/praveen-ilangovan/pensieve/claude.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 | $0.00099 | $0.02665 |
| Opus 5 | $0.00049 | $0.01333 |
| Sonnet 5 | $0.00020 | $0.00533 |
| Haiku 4.5 | $0.00010 | $0.00266 |
Grade A, and why
pensieve 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 11d 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 — 157 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Pensieve
Pensieve is the user's manually-triggered memory — an information lake that groups itself. Four things:
- stream — a top-level domain (
recs,employment). - thread — a focused sub-topic inside a stream (e.g. a person who recurs).
- note — an atomic piece of information (the unit you capture).
- entity — a named thing notes refer to (a person/org/topic). Notes get tagged with the entities they mention; once an entity recurs enough, it can earn its own thread.
Operate it through the pensieve MCP tools. You judge; the engine writes — never
edit the store directly.
Vocabulary
- Speak in streams, threads, notes, and the real names of things ("Rafia", "the meeting"). Never expose storage mechanics — node, edge, attachment, tag (if a tool error leaks "node", say "stream"/"thread").
capture/fetchare your flows (judgment). The MCP tools are mechanical ops you compose; callingadd_noteisn't capturing — you route, resolve, and get the user's OK around it.
Tools
list_streams()— the stream index (load before routing).create_stream(name, purpose)/edit_stream(stream, name?, purpose?)— a new domain (deliberate; confirm first) / rename or repurpose one (id is immutable).list_entities()— the entity registry with note counts +promotableflag.find_entities(query)— fuzzy search the registry (resolve a mention; recall).add_note(stream, text, entities=[…])— record a note + tag the entities it mentions.tag_note(note, entities=[…])/untag_note(note, entity)— link / unlink an entity on an existing note (fix tagging).edit_note(note, text)— fix a genuine mistake (a change in the world is a new note).get_stream(stream)— a stream's (or thread's) view.get_entity(entity)/edit_entity(entity, name?, aliases?)— recall everything about an entity / rename it.promote_entity(entity, stream)— give a recurring entity its own thread.add_asset(target, location, hint?, label?, kind?)/list_assets(target)/remove_asset(asset)— attach/list/remove a by-reference pointer (repo/file/dir/URL/ image/doc) on a stream, thread, or note. Pointers only — see ASSETS below.- Remove / restore — all soft & reversible (tell the user so):
remove_stream/remove_note/remove_entity, undone byrestore_stream/restore_note/restore_entity.remove_entityunlinks — it never deletes a note.
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.
- 11d ago First seen · 157 lines · 99 tokens per session scan A d0ab9e598f7e
pensieve is a skill published in the GitHub repository praveen-ilangovan/pensieve (0 stars, last pushed 12d ago), licensed MIT. It adds 99 tokens to every session and 2,665 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.
Other skills, from other repositories
add-mwe-memory
Give this NanoClaw's agents persistent, governed memory from an mwe-mcp server. Every turn is remembered and recalled mechanically, per speaker, with the built-in memory tree switched off and no session carried between turns.
mwe-mcp-memory
Use this whenever you are connected to the user's mwe-mcp memory server (the "mwe-mcp" MCP connector). It tells you how to use that memory well — recall what you've discussed before, keep a running chronicle of your conversations, and capture facts as you go, without bothering the user.
plur-create-engrams
Create or improve PLUR engrams from conversations, documents, decisions, observations, and explicit preferences. Use for memory extraction, engram authoring, or reviewing proposed memories, including global, scoped, pinned, retrieved, and provisional knowledge. Ordinary use of existing memories does not require this…
plur-memory
Persistent learning for AI agents. Open engram format. Your agent learns from corrections, remembers across sessions, and transfers knowledge across domains.
lemmalog
Externalize working memory and logical state into the lemmalog Datalog engine (MCP). Use for ANY multi-step task where state should outlive one context window or span agents: long investigations, debugging sessions, audits, multi-agent searches, systematic explorations, planning with many interdependent constraints…
plur-session-end
Extract durable learnings at the end of a session. Saves corrections, preferences, and codebase patterns as engrams — nothing ephemeral, nothing sensitive.