From xlfg-engineering
Use docs/xlfg and .xlfg as the file-based context system for lean, autonomous SDLC runs.
npx claudepluginhub flrngel/xlfg --plugin xlfg-engineeringThis skill uses the workspace's default tool permissions.
Use file-based context to keep long-horizon work reliable and legible.
Uses filesystem for agent context overflow: persists tool outputs/state, enables dynamic discovery/scratch pads, and selective retrieval to fix context bloat and retrieval issues.
Provides filesystem-based context management for AI agents, enabling dynamic discovery, just-in-time file loading, and reduced context window bloat. Use for state persistence, scratch pads, and sub-agent info sharing.
Designs three-tier context hierarchy (CLAUDE.md, priming commands, on-demand reads) for progressive loading and optimal management in Claude Code projects.
Share bugs, ideas, or general feedback.
Use file-based context to keep long-horizon work reliable and legible.
Treat files as the system of record:
docs/xlfg/knowledge/ + docs/xlfg/meta.jsondocs/xlfg/knowledge/current-state.mddocs/xlfg/runs/.xlfg/docs/xlfg/
index.md
meta.json
knowledge/
current-state.md
quality-bar.md
decision-log.md
patterns.md
testing.md
ux-flows.md
failure-memory.md
harness-rules.md
ledger.jsonl
commands.json
agent-memory/
...
migrations/
<from>-to-<to>.md
runs/
<run-id>/
context.md
memory-recall.md
spec.md
test-contract.md
test-readiness.md
workboard.md
research.md # optional
diagnosis.md # optional
solution-decision.md # optional
flow-spec.md # optional
env-plan.md # optional
proof-map.md # optional
risk.md # optional
verification.md
verify-fix-plan.md
review-summary.md # optional
run-summary.md # optional
compound-summary.md
current-state-candidate.md # branch-local when needed
tasks/
<task-id>/
task-brief.md
.xlfg/
runs/
<run-id>/
doctor/
<ts>/
verify/
<ts>/
spec.md current instead of copying the same truth into multiple planning files.workboard.md carry stage truth and verification.md carry proof evidence.