Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add stark-ai-de/agent-skills --skill claude-memory-curatorgit clone --depth 1 https://github.com/stark-ai-de/agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/stark-ai-de/agent-skills/claude-memory-curator)<a href="https://agentmods.dev/skills/stark-ai-de/agent-skills/claude-memory-curator"><img src="https://agentmods.dev/badge/skills/stark-ai-de/agent-skills/claude-memory-curator/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/stark-ai-de/agent-skills/claude-memory-curator"><img src="https://agentmods.dev/badge/skills/stark-ai-de/agent-skills/claude-memory-curator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Memory Poisoning · line 159 Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.Fix: Protect agent memory and state from modification by untrusted content. Use read-only memory for critical instructions and validate all state changes.
- medium Prompt Injection · line 34 Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.Fix: Remove the large whitespace padding (blank-line blocks or long space runs) and review any content hidden below or to the right of it. Keep skill files compact and reviewable so no instructions can be
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00111 | $0.03391 |
| Opus 5 | $0.00056 | $0.01695 |
| Sonnet 5 | $0.00022 | $0.00678 |
| Haiku 4.5 | $0.00011 | $0.00339 |
Grade A, and why
claude-memory-curator scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 183 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Claude Memory Curator
Goal
Audit Claude Code durable context as user-owned agent state: expose stale, unsafe, duplicated, over-broad, conflicting, misplaced, or unenforceable entries; propose better destinations; and route review, planning, persistence, and cleanup through one explicit contract.
Keep the subject scoped to Claude Code surfaces: CLAUDE.md, CLAUDE.local.md, .claude/rules/, user-level Claude rules, settings, hooks, managed policy evidence, and auto memory files. Do not treat this as a Codex, Cursor, Claude app, or Anthropic API Memory Stores curator.
When to use
- Use for review, placement, planning, or cleanup of the Claude Code durable surfaces named in the description.
- Use when those surfaces are stale, conflicting, sensitive, over-broad, ignored, or causing Claude to reuse old guidance.
When not to use
- Do not use for Codex, Cursor, Claude web/app, Anthropic API Memory Stores, generic prompt engineering, or repo-doc cleanup without Claude durable context.
- Keep review requests read-only. Treat managed policy, including
/etc/claude-code/CLAUDE.mdandmanaged-settings.json, as read-only unless exact edit authority is explicit.
Workflow selection
Always expose these workflows in this order. plan-run-cleanup-file is always first and Recommended:
| Workflow | Delivery | Result |
|---|---|---|
plan-run-cleanup-file (Recommended) |
One redacted file record | Full review, user-approved cleanup plan, backup, execution, and verification. |
review-chat |
Chat only | Full read-only review and recommendations. |
review-file |
One redacted file record | Full read-only review and recommendations. |
cleanup-chat |
Chat plus backup | Full review followed by direct high-confidence atomic cleanup and verification. |
cleanup-file |
One redacted file record plus backup | Persist the review, then directly apply high-confidence atomic cleanup and verification. |
plan-cleanup-chat |
Chat only | Full review and user-approved cleanup plan; no cleanup. |
plan-cleanup-file |
One redacted file record | Full review and user-approved cleanup plan; no cleanup. |
plan-run-cleanup-chat |
Chat plus backup | Full review, user-approved cleanup plan, backup, execution, and verification. |
What ships with it
11 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- assets/cleanup-plan-template.md 1.9 KB
- assets/review-report-template.md 2.3 KB
- references/classification-rubric.md 5.2 KB
- references/conflict-resolution.md 1.8 KB
- references/context-surface-anatomy.md 2.8 KB
- references/example-review-report.md 2.3 KB
- references/safe-editing-procedure.md 5.3 KB
- references/settings-and-hooks.md 1.6 KB
- scripts/backup-claude-memory.mjs 42 KB runs code
- scripts/inventory-claude-memory.mjs 13 KB runs code
- scripts/scan-claude-memory-risks.mjs 13 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 183 lines · 111 tokens per session scan A ed892c5a8db7
claude-memory-curator is a skill published in the GitHub repository stark-ai-de/agent-skills (5 stars, last pushed today), licensed Apache-2.0. It adds 111 tokens to every session and 3,391 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
levelup-specify
Extract Context Directive Records (CDRs) from the current session after completing work. Identifies reusable patterns (rules, personas, examples, evals) and captures directive compliance cases for team-ai-directives.
change-publish
Promote accepted Change Decision Records (ChDRs) from drafts to project memory at .adlc/memory/chdr/, write OKF-style frontmatter, and regenerate the boot-facing .adlc/memory/chdr.md index that team-boot injects at session start. Use after /change-clarify has accepted ChDRs.
captain-recall
Use when the user asks WHY or WHAT-HAPPENED questions about this system — historical decisions, past incidents, why code/config is shaped a certain way, where something is deployed, who decided what, or the state of work in flight. Queries the tsubasa knowledge graph for cited answers.
captain-inject
Use when the user STATES A FACT about the system or tells the captain to remember/learn something — "update your knowledge", "remember this", "we dropped X for Y", "the outage was caused by Z", environment URLs, team contacts, deployment flows, tribal knowledge, corrections to what the Captain believes. Validates…
decision
Single entry point for all decision operations: capture, search, manage. Routes based on natural language intent.
capture-mcp-context
Record durable context from MCP servers and design or project tools into repository memory as proposed. Use when working with an MCP server like Figma, Linear, Jira, or Sentry, or after pulling design, ticket, or error context, to persist it for future sessions.