
Palace Index Curator
- 24 installs
- 325 repo stars
- Updated August 2, 2026
- athola/claude-night-market
Keep a growing Memory Palace web-capture index from becoming a write-only backlog by reporting health, promoting entries, and surfacing relevant captures during agent sessions.
About
Palace Index Curator is a governance skill for the Memory Palace web-capture pipeline. Night Market hooks append every WebFetch and WebSearch to hooks/memory-palace-index.yaml as markdown plus index rows, but defaults leave most entries pending seedlings that never graduate. Solo and indie builders who rely on captured research during long agent sessions install this skill to drain the backlog on a schedule: run analytics for corpus health, promote entries with a dry-run gate, and optionally attach SessionStart surfacing so prior fetches reappear in context. It is not for one-off ingestion or ad hoc search—those stay on knowledge-intake and normal retrieval. Think of it as the janitorial layer between passive capture and a corpus you actually trust when scoping, building, or iterating.
- Read-only corpus health report: inert ratio, orphans, and topic clusters
- Dry-run-first promotion engine that advances pending/seedling captures using decay_model, keyword_index, and marginal_va
- SessionStart hook to surface stored web research during live agent work
- Targets default backlog states: routing_type pending, maturity seedling, importance_score 50
- Complements auto-capture from WebFetch/WebSearch into hooks/memory-palace-index.yaml
Palace Index Curator by the numbers
- 24 all-time installs (skills.sh)
- Ranked #1,974 of 3,282 Productivity & Planning skills by installs in the Skillselion catalog
- Data as of Aug 5, 2026 (Skillselion catalog sync)
npx skills add https://github.com/athola/claude-night-market --skill palace-index-curatorAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Installs | 24 |
|---|---|
| repo stars | ★ 325 |
| Last updated | August 2, 2026 |
| Repository | athola/claude-night-market ↗ |
What it does
Keep a growing Memory Palace web-capture index from becoming a write-only backlog by reporting health, promoting entries, and surfacing relevant captures during agent sessions.
Files
Palace Index Curator
Overview
The web-research hooks auto-capture every WebFetch and WebSearch into hooks/memory-palace-index.yaml, storing each as a markdown file and an index entry. Captures land at the defaults routing_type: pending, maturity: seedling, importance_score: 50, and nothing advances them. Left alone, the index becomes a write-only graveyard: the majority of entries are never incorporated, analyzed, or surfaced.
This skill drains that backlog and keeps it drained. It wires the capture index to the corpus tooling the plugin already ships (decay_model, keyword_index, marginal_value) through three commands: a read-only report, a dry-run-first promotion engine, and a SessionStart surfacing hook.
When to Use
- The capture backlog has grown and most entries are still
pending. - You want a corpus health report (inert ratio, orphans, topic clusters).
- You want stored research surfaced automatically during sessions.
When NOT to Use
- Ingesting a single new resource: use
knowledge-intake. - Searching stored knowledge ad hoc: use
knowledge-locator. - Tending a digital garden file: use
digital-garden-cultivator.
Workflow
1. Analyze (read-only)
uv run python scripts/memory_palace_cli.py index reportReports total entries, the inert ratio, orphaned captures (entries whose backing file is gone), the largest topic clusters by domain, and the top promotion candidates. Writes nothing.
2. Incorporate (dry-run, then apply)
# Dry run: prints promote/archive proposals, writes nothing.
uv run python scripts/memory_palace_cli.py index promote
# Apply: backs up the index under data/backups/, then persists.
uv run python scripts/memory_palace_cli.py index promote --applyEach pending entry is classified into one action:
- promote: recent, authoritative, or clustered. Gets a real
importance score, a routing type, and maturity seedling -> growing.
- archive: orphaned or older than the archive horizon and never
revisited. Marked archived rather than promoted, following the principle that unused captures should drain, not accumulate.
- hold: everything else stays
pendingwith no change.
Applying is idempotent: promoted and archived entries are no longer pending, so a second run proposes nothing new. The dry-run diff is always shown before --apply writes.
3. Surface (learn)
A SessionStart hook (hooks/index_surfacer.py) names the highest-value promoted captures at the start of a session. It is disabled by default. Enable it in memory-palace-config.yaml:
feature_flags:
context_injection: trueThe hook only speaks when promoted entries clear the importance floor, and it exits silently on any error so it can never block a session.
Design Notes
- Promotion uses only structural signals (recency, domain authority,
cluster size). The decision logic is deterministic; no model call gates a transition.
- The decay half-lives (14/30/90 days) are tunable priors, not retention
constants. Wixted & Ebbesen (1997) and Murre & Dros (2015) show forgetting follows a power law; FSRS (Ye, Su & Cao, 2022) validates exponential decay only with a learned per-item half-life. Calibrate against reopen logs if usage data accrues.
- Retrieval stays keyword-first (
cache_lookup/keyword_index);
embeddings are not required at the current corpus scale. BM25 is the workhorse up to ~5000 documents; embeddings add value only for vocabulary-mismatch discovery.
- Near-duplicate detection layers SHA-256 exact match (present via
content_hash) then MinHash with k-shingling for near-duplicates (Broder, 1997). SimHash is preferable only at tens of thousands of documents.
- Importance formula: `relevance = w1 centrality + w2 decay(t) +
w3 * usage. The plugin ships all three terms (graph_analyzer PageRank, decay_model, usage_tracker`).
Exit Criteria
- [ ]
index reportruns and prints the inert ratio and orphan count
for the live index.
- [ ]
index promote(no flag) prints proposals and writes nothing
(the index file is byte-identical afterward).
- [ ]
index promote --applycreates a timestamped backup under
data/backups/ before persisting, and a re-run proposes nothing.
- [ ] With
context_injection: true, a SessionStart event surfaces the
top promoted captures; with the flag off, it stays silent.
- [ ] Failure modes (missing index, corrupt YAML, missing backing files)
are handled without raising: report degrades, promote holds, hook exits silently.