Dorveille recall

Skill sb-arnav/dorveille/plugins/dorveille/skills/dorveille-recall

The trust layer for AI agent self-improvement, native to Claude Code: cross-session memory, self-forged skills, and FTS5 episodic recall — but every lesson carries provenance + a trust score, is challenged by an independent verifier, and re-checked by a falsifiable probe. On your own subscription.

Install

npx -y skills add sb-arnav/dorveille --skill dorveille-recall

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use to recall something from a PAST Claude Code session — when the user says "did we discuss", "what did I say about", "last time", "a while back", "you helped me with X before", "remember when", or references earlier work not in the current context. Searches every past session transcript (FTS5, zero-LLM) and returns the matching conversation with enough surrounding context to act on.

SKILL.md

2.5 KB, 510 tokens by cl100k_base, as published. Nobody here has run it

dorveille-recall — search your past sessions

Claude Code keeps a full transcript of every session, but they're invisible to the agent mid-conversation. dorveille-recall makes them searchable, so you can pull back a decision, a config value, or a discussion from weeks ago instead of asking the user to repeat themselves. This is the Hermes "it remembered a detail from three weeks ago" capability, native to Claude Code.

Reach for this BEFORE saying "I don't have that context" or asking the user to re-explain prior work. If the user references something that plausibly happened in an earlier session, search first.

How to use it

dorveille recall search "<keywords>"           # find sessions; returns snippets + session ids
dorveille recall scroll <session> <anchor>     # widen context around a hit (anchor = uuid or line no shown in search)
dorveille recall browse                        # most recent sessions, newest first
dorveille recall index                         # refresh the index (auto-runs on SessionEnd; safe to run anytime)

(If dorveille isn't on PATH: python3 ${CLAUDE_PLUGIN_ROOT}/bin/dorveille-recall ....)

Workflow

Search with the strongest distinctive keywords from the user's reference (project names, error strings, tool names, decisions). Results are ranked sessions, each with highlighted ⟦match⟧ snippets and a session id.
If a snippet looks right but you need more, scroll around its anchor to read the surrounding turns (--window N for a wider view).
Use what you find to answer directly — cite that it's from a past session and roughly when, so the user knows you recalled it rather than guessed.

Notes

Keyword/substring search (FTS5), not semantic — vary terms if the first query misses. Bare terms are OR'd, so extra keywords broaden rather than narrow.
Indexes only human-readable conversation (user messages + assistant text), so results are signal, not tool spew.
The index is local at ~/.dorveille/sessions.db and refreshes incrementally; if a very recent session is missing, run dorveille recall index.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.