Skill

xlfg-recall-phase

Internal xlfg phase skill. Use only during /xlfg runs to perform deterministic recall and record the smallest relevant carry-forward memory.

npx claudepluginhub flrngel/xlfg --plugin xlfg-engineering

Tool Access

This skill is limited to using the following tools:

ReadGrepGlobLSBashEditWrite

Preview

Use only during `/xlfg` orchestration.

SKILL.md

Similar Skills

harness-recall

Classifies user recall intents (resume, decisions, prior issues, sessions, keywords) into 5 routes, invoking harness-mem tools or SSOT markdown files to deliver source-cited 1-line summaries.

harness-mem

hipocampus-recall

149

Retrieves project memory for user queries via three-step fallback: ROOT.md topics triage, frontmatter manifest LLM selection from weekly/monthly notes, qmd search. Use when past knowledge may apply.

hipocampus

memory-skills

Implements CoALA memory architecture for Claude: stores working, episodic, semantic, procedural memories in filesystem for recall, reflection, learning across sessions.

17 files

cowork-skills

Stats

Parent Repo Stars1

Parent Repo Forks1

Last CommitMar 25, 2026

Actions

View Source View Plugin View on GitHub View README

Help us improve

Share bugs, ideas, or general feedback.

xlfg-recall-phase

Use only during /xlfg orchestration.

Input: $ARGUMENTS (RUN_ID or latest)

Objective

Perform deterministic recall before broad repo scanning and record only the smallest relevant carry-forward context.

Process

Resolve RUN_ID, DOCS_RUN_DIR, and DX_RUN_DIR.

Read in this order when present:

docs/xlfg/knowledge/current-state.md
the latest related run if you are continuing work
docs/xlfg/knowledge/*.md
docs/xlfg/knowledge/agent-memory/*.md
docs/xlfg/knowledge/ledger.jsonl

Prefer the helper when available:

xlfg recall yesterday
xlfg recall '<quoted lexical query>'
typed lexical recall when the failure or flow is known

Write memory-recall.md with:

sources checked
strong matches
carry-forward rules for this run
rejected near-matches
an explicit no-hit statement when nothing relevant applies

Update spec.md so the recall summary is visible in the single source of truth.

Update workboard.md so recall is DONE and planning is the next active step.

Guardrails

Do not write vague history dumps.

Do not invent semantic recall when lexical evidence is weak.

Keep the memory note compact enough to help the next phase immediately.