Skip to content
Back to skills

Stale Pending Memory Guard

ASecurity

Before reporting any memory hit containing "pending/awaiting/needs/consider" as current fact, verify it against the actual source. Prevents propagating resolved items that memory hasn't caught up to.

  • 3 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 1, 2026
ai-agentspythonrustgit

Security analysis

A100/100

Scanned October 1, 2026

npx -y skills add williamblair333/Uncle-J-s-Refinery --skill stale-pending-memory-guard --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Stale Pending Memory Guard?

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

Security grade badge for Stale Pending Memory Guard
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/williamblair333-stale-pending-memory-guard/badge)](https://www.skillsdirectory.com/skills/williamblair333-stale-pending-memory-guard)

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: stale-pending-memory-guard
description: Before reporting any memory hit containing "pending/awaiting/needs/consider" as current fact, verify it against the actual source. Prevents propagating resolved items that memory hasn't caught up to.
---

## When to use

Run this guard inside `prior-art-check` whenever a memweave (`mw_search.py`) hit's body contains action-pending language. Without it, a stale "awaiting review" entry gets reported as current fact even after the upstream work completed — as happened with an upstream PR whose note said "awaiting review" after it had already merged.

This guard fills a gap that healthcheck 9e does **not** cover: 9e catches technical staleness (are we behind upstream?), not tracking staleness (does our note still match reality?).

## Trigger words

Flag any memory hit whose body contains: `pending`, `awaiting`, `needs`, `consider`, `TODO`, `blocked`, `in progress`, `not yet`, `will be`, `outstanding`.

## Steps

1. **Surface hits normally** via memweave (`.venv-memweave/bin/python scripts/memweave/mw_search.py "<query>" --k 5`).
2. **Scan each hit's body** for trigger words (case-insensitive).
3. **Verify flagged entries** before reporting:
   - PR/issue → `gh pr view <number>` or `gh issue view <number>`
   - Package/code → freshness check or `git log --oneline -5`
   - Config/infra → read the live file or service status
4. **Report the verified state**, not the memory text.
5. **Update stale entries** by editing the corresponding note in the memweave corpus (`~/.uncle-j-memory/memory/`) with corrected status + date; the nightly `sync_memory.sh` re-embeds it.
6. Continue with the rest of `prior-art-check` normally.

## What NOT to flag

- Entries describing completed past facts: "was fixed in v3.3.5", "merged 2026-05-15" — historical, not pending.
- Entries where "pending" appears in a key name but doesn't imply outstanding action.

## Why this matters

Memory records are hypotheses, not facts. Any entry written when work was in-flight can silently become wrong after the work completes. Propagating a stale "pending" status wastes a session turn and erodes trust in memory as a source.

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.