Cortex — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Cortex (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Local-first AI memory layer with hybrid vector + BM25 retrieval. No API keys. Stores memories in LanceDB at ~/.cortex/lance_db/.
cortex save "content" --namespace <ns> --type <type> -i <0-1> --tags "tag1,tag2" --project <name>cortex search "query" --namespace <ns> --type <type> -n <limit> --min-importance <n> --project <name>Hybrid scoring: 35% vector similarity + 30% BM25 + 15% recency + 10% importance + 5% access frequency + 5% type boost.
cortex import file.md --namespace <ns> --smart --no-dedup
cortex ingest ~/folder --recursive --namespace <ns> --smart --ext .md,.txt--smart extracts lines prefixed with Decision:, Lesson:, Bug: etc.
cortex status # Overview with namespace distribution
cortex health # Brain health metrics
cortex audit # Find duplicates, stale memories (slow for large DBs)
cortex decay --dry-run # Preview importance decay
cortex decay --apply --half-life 30 # Apply decay (30-day half-life)
cortex consolidate --dry-run # Find similar memory clusters
cortex consolidate --apply # Merge clusters into summaries
cortex curate --auto # Delete low-value memories
cortex export --namespace health # Export filtered as markdownmemory_save — save with namespace, type, importance, tags, metadatamemory_search — hybrid search with namespace/type filtersmemory_context — stats overviewmemory_forget — delete by IDmemory_reflect — top accessed + highest importance| Namespace | What goes here | Examples |
|---|---|---|
health | Food, workouts, sleep, body metrics | "Ran 5k in 24:30", "Slept 7.5 hours" |
projects/<name> | Per-project memories | "projects/myapp", "projects/kalshi" |
personal | Relationships, reflections, plans | "Feeling burnt out this week" |
daily | Raw daily logs | "2026-03-04: shipped namespace feature" |
learnings | Mistakes, corrections, patterns | "Never trust sub-agent success claims" |
people | Info about specific people | "Sarah prefers async communication" |
general | Default / uncategorized | Anything that doesn't fit above |
general. cortex decay --apply
cortex curate --auto
cortex health cortex consolidate --dry-run # review first
cortex consolidate --applyhealthprojects/<name>peoplelearningsgeneralscore = 0.35×vector + 0.30×bm25 + 0.15×recency + 0.10×importance + 0.05×access_freq + 0.05×type_boostMemories accessed more often score higher. Unaccessed memories decay over time.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.