npx claudepluginhub mistakeknot/interagency-marketplace --plugin intermemThis skill uses the workspace's default tool permissions.
Guides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.
Migrates code, prompts, and API calls from Claude Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5, updating model strings on Anthropic, AWS, GCP, Azure platforms.
Reviews prose for communication issues impeding comprehension, outputs minimal fixes in a three-column table per Microsoft Writing Style Guide. Useful for 'review prose' or 'improve prose' requests.
Scan auto-memory for the current project, identify stable facts, and promote them to AGENTS.md/CLAUDE.md with user approval.
Run the intermem synthesis pipeline using the Python library:
.intermem/stability.jsonl — only promote entries stable across 3+ snapshotsFind the intermem plugin install path and run from there:
INTERMEM_DIR="$(find ~/.claude/plugins/cache -path '*/intermem/*/pyproject.toml' -not -path '*/tests/*' -printf '%h\n' 2>/dev/null | sort -V | tail -1)"
cd "$INTERMEM_DIR" && uv run python -m intermem --project-dir /path/to/project --json
Replace /path/to/project with the actual project directory (usually the current working directory before cd).
Add --validate to enable citation checking. Entries referencing deleted files/modules get filtered:
cd "$INTERMEM_DIR" && uv run python -m intermem --project-dir /path/to/project --validate --json
Additional flags:
--validate — Enable citation validation in the synthesis pipeline--no-validate — Skip citation validation (Phase 0.5 behavior)--validate-only — Standalone mode: check already-promoted entries and report stale ones--project-root /path — Override project root for citation resolution (default: --project-dir value)--max-candidates N — Cap candidates per run (default 10, 0=unlimited). Deferred entries promote on subsequent runsOn first run (no .intermem/stability.jsonl), the scanner records a baseline snapshot. All entries are scored "recent" — none are promoted. The user must run synthesis again after a few more sessions to build history.
Display: "Building baseline — recorded N entries. Run again after your next few sessions to identify stable facts."
Present candidates as a numbered list grouped by target document:
Candidates for AGENTS.md:
1. [stable] Oracle CLI requires DISPLAY=:99 (section: Cross-Cutting Lessons)
2. [stable] Never use > file redirect with oracle (section: Cross-Cutting Lessons)
3. [similar to existing] Git credential lock fix... (section: Cross-Cutting Lessons)
Candidates for CLAUDE.md:
4. [stable] Always use uv run for Python deps (section: Tool Usage)
Enter selections (e.g., "approve 1,2,4 / reject 3" or "all"):