Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/richfrem/agent-plugins-skillsnpx agentmods add skills/richfrem/agent-plugins-skills/managing-rlm-cacheWrote 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/richfrem/agent-plugins-skills/managing-rlm-cache)<a href="https://agentmods.dev/skills/richfrem/agent-plugins-skills/managing-rlm-cache"><img src="https://agentmods.dev/badge/skills/richfrem/agent-plugins-skills/managing-rlm-cache/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/richfrem/agent-plugins-skills/managing-rlm-cache"><img src="https://agentmods.dev/badge/skills/richfrem/agent-plugins-skills/managing-rlm-cache.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00075 | $0.01073 |
| Opus 5.5 | $0.00030 | $0.00429 |
| Sonnet 5.5 | $0.00015 | $0.00215 |
| Haiku 4.5 | $0.00007 | $0.00107 |
Grade A, and why
managing-rlm-cache 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 2d 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 — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Managing RLM Cache
Read the matching workflow before acting. Run commands from this skill's folder;
set AGENT_MEMORY_PROJECT_ROOT to the consumer project, including for global installs.
Contents
Interaction · Constraints · Workflow · Quick start · Verification
Interaction
Treat slash-invocation arguments as supplied requirements; do not ask for them again.
For setup, semantic-cache onboarding or layer upgrades, read guided setup. Conduct it here in the main conversation; reuse known answers and ask one unresolved question per turn. Offer available standalone/combined choices, preview the config changes, then apply the authorized scope. Continue other skills' selected modes in the same session; preserve answers and never launch a setup agent. For an explicit query/audit request, use the matching mode directly without restarting an interview.
Constraints
- Read each source fully before summarizing. Report source paths and injected summaries.
- Never edit ledger files manually: use
scripts/inject_summary.pyorscripts/swarm_run.py. - Audit and query are read-only. Show cleanup dry runs and confirm scope before applying deletions.
- Use configured profiles in the consumer's
.agent/learning/rlm_profiles.json. - Search summaries first, then delegate semantic retrieval to managing-vector-db if installed, then use scoped exact search. Do not import another skill's Python implementation.
- Ledger lookup/injection require only Python's standard library. Batch jobs use PyYAML; model-backed distillation additionally needs the selected CLI or configured service.
Workflow
| Request | Read first |
|---|---|
| Initialize, add profile or choose stack | Guided setup, initialization, bootstrap guide |
| Keyword lookup / three-phase search | Search |
| Read sources and inject summaries | Distillation |
| Maintain coverage / batch work | Curation |
| Compare manifest with cache | Audit |
| Preview and prune stale entries | Cleanup |
| Wiki summary, bullets and deep layers | Use managing-obsidian-vaults, wiki distillation mode; verify it is installed |
What ships with it
60 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/examples/rlm_profiles.json 636 B
- assets/examples/rlm_summary_cache_manifest.json 153 B
- assets/examples/rlm_tools_manifest.json 158 B
- assets/jobs/distill_wiki.job.md 1.7 KB
- assets/templates/distiller_manifest.json 130 B
- assets/templates/manifest-index.json 415 B
- assets/templates/prompts/rlm/rlm_summarize_general.md 1.4 KB
- assets/templates/prompts/rlm/rlm_summarize_tool.md 1.8 KB
- assets/templates/rlm_manifest.json 368 B
- assets/templates/rlm_summary.md.j2 79 B
- evals/evals.json 16 KB
- evals/legacy/rlm-audit/evals.json 877 B
- evals/legacy/rlm-audit/results.tsv 34 B
- evals/legacy/rlm-cleanup-agent/evals.json 964 B
- evals/legacy/rlm-cleanup-agent/results.tsv 172 B
- evals/legacy/rlm-curator/evals.json 948 B
- evals/legacy/rlm-curator/results.tsv 301 B
- evals/legacy/rlm-distill-agent/evals.json 1004 B
- evals/legacy/rlm-distill-agent/results.tsv 172 B
- evals/legacy/rlm-init/evals.json 693 B
- evals/legacy/rlm-init/results.tsv 172 B
- evals/legacy/rlm-search/evals.json 1.2 KB
- evals/legacy/rlm-search/results.tsv 172 B
- evals/README.md 569 B
- evals/results.tsv 73 B
- evals/task-success.json 2.1 KB
- references/acceptance-criteria.md 2.6 KB
- references/BLUEPRINT.md 360 B
- references/cheapest_models.json 845 B
- references/cheapest_models.md 1.1 KB
- references/diagrams/distillation_process.mmd 919 B
- references/diagrams/logic.mmd 1002 B
- references/diagrams/rlm_late_binding_flow.mmd 1.1 KB
- references/diagrams/rlm_mechanism_workflow.mmd 2.0 KB
- references/diagrams/rlm_tool_enrichment_flow.mmd 819 B
- references/diagrams/rlm-factory-architecture.mmd 615 B
- references/diagrams/rlm-factory-dual-path.mmd 1019 B
- references/diagrams/rlm-factory-workflow.mmd 2.3 KB
- references/diagrams/search_process.mmd 2.0 KB
- references/diagrams/unpacking.mmd 561 B
- references/diagrams/workflow.mmd 578 B
- references/gap_analysis_rlm_v1.md 370 B
- references/interactive-setup.md 5.6 KB
- references/logic.mmd 1001 B
- references/prompt.md 357 B
- references/research-summary.md 367 B
- references/research/2512.24601v1.pdf 8057 KB
- references/research/summary.md 3.5 KB
- references/resource-index.md 4.5 KB
- references/RLM_ARCHITECTURE.md 367 B
- references/rlm-audit.md 1.9 KB
- references/rlm-bootstrap-guide.md 3.8 KB
- references/rlm-cleanup-agent-acceptance-criteria.md 306 B
- references/rlm-cleanup-agent.md 2.1 KB
- references/rlm-curator-acceptance-criteria.md 285 B
- references/rlm-curator-fallback-tree.md 1.4 KB
- references/rlm-curator.md 2.7 KB
- references/rlm-distill-agent-acceptance-criteria.md 298 B
- references/rlm-distill-agent.md 2.9 KB
- references/rlm-init-acceptance-criteria.md 440 B
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.
- 2d ago First seen · 77 lines · 75 tokens per session scan A c72362ad8656
managing-rlm-cache is a skill published in the GitHub repository richfrem/agent-plugins-skills (7 stars, last pushed yesterday), licensed MIT. It adds 75 tokens to every session and 1,073 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-10-24.
Other skills, from other repositories
ap-sweeper
Report the compatibility redirect to ap-independent-checker; this retired role cannot perform new work.
knowledge-graph-management
Capture, validate, query, and sync architectural patterns and design decisions in the knowledge graph.
context-recovery
Automatically recover working context after session compaction or when continuation is implied but context is missing. Works across Discord, Slack, Telegram, Signal, and other supported channels.
iwe-restore
A recovery skill that restores an agent’s saved memory and instruction files from an exocortex backup when moving to a new computer or workspace. It shows what it found and asks for confirmation before copying files.
a2wave-memory
Progressively recall and maintain a2wave cross-session memory through a compact startup catalog, bounded topics, and searchable history.
memory-patterns
Persistent memory patterns for moflo agents — session memory, long-term knowledge, pattern learning, and cross-session context via moflo's node:sqlite + HNSW vector store. Use when building stateful agents or assistants that need to remember across runs.