Skip to content
Back to skills

Np Core Recall

ASecurity

Pull episodic working-memory themes ("what we did / decided / where we left off") from nervepack's episodic/ layer on demand. Use when the user says "/recall", "what did we do on X", "remind me where we left off on Y", or wants prior-session context that wasn't auto-injected. Read-only — episodic is the lowest-authority layer; durable skills/sources/wiki override it.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 19, 2026
ai-agentspythonbash

Works with

  • cli

Security analysis

A100/100

Scanned September 19, 2026

npx -y skills add pat-browne/nervepack --skill np-core-recall --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Np Core Recall?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Np Core Recall
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/pat-browne-np-core-recall/badge)](https://www.skillsdirectory.com/skills/pat-browne-np-core-recall)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: np-core-recall
description: Pull episodic working-memory themes ("what we did / decided / where we left off") from nervepack's episodic/ layer on demand. Use when the user says "/recall", "what did we do on X", "remind me where we left off on Y", or wants prior-session context that wasn't auto-injected. Read-only — episodic is the lowest-authority layer; durable skills/sources/wiki override it.
---

# recall

The on-demand counterpart to the automatic `UserPromptSubmit` injection
(`engine/nervepack_engine/hooks/episodic_recall.py`, dispatched via `cli.py hook
episodic-recall`). Auto-injection only fires on a session's first
couple of prompts and only on keyword hits; `/recall` lets you pull any theme,
any time.

## Trigger phrases
- `/recall`, `/recall <topic-or-keywords>`
- "what did we do on X", "where did we leave off on Y"
- "pull up our prior context about Z"

## Steps

### 1. Locate candidate themes

Episodic lives in the **content overlay** under `memory/episodic/`, not the
engine repo. Resolve the roots (merge-aware, team > personal):

```bash
python3 "${NP_DIR:-$HOME/Code/nervepack}/engine/nervepack_engine/np_content.py" layer_roots episodic  # one dir per overlay root
```

For each root: `ls <root>/*.md` (available themes) and `sed -n '1,80p'
<root>/INDEX.md` (topic | keywords index). Match the user's query against topic
slugs and the `keywords` column. If they gave no query, list the most recently
updated themes and ask which.

### 2. Read the matching theme(s)
Read `<root>/<topic>.md` for the best match(es). Prefer the most recent
entries; mention the `## Rolled-up summary` block if older context is relevant.

### 3. Present
Summarize what was done / decided / where it was left off, citing the file
path. **Flag the authority level**: episodic context is working narrative and
may be stale — if it conflicts with a `skills/`, `sources/`, or `wiki/` page,
the durable layer wins. Say so when relevant.

## What this skill does NOT do
- Does not write to the episodic layer (capture + the episodic-maintain agent own writes).
- Does not promote anything to `skills/` — if a recalled entry is actually a
  durable rule, route it through [[np-core-contribute]].
- Does not surface secrets; if a theme somehow contains one, redact in your
  reply and note it for cleanup.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.