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 lukacf/meerkat --skill session-managementgit clone --depth 1 https://github.com/lukacf/meerkatWrote 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/lukacf/meerkat/session-management)<a href="https://agentmods.dev/skills/lukacf/meerkat/session-management"><img src="https://agentmods.dev/badge/skills/lukacf/meerkat/session-management/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/lukacf/meerkat/session-management"><img src="https://agentmods.dev/badge/skills/lukacf/meerkat/session-management.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00016 | $0.00405 |
| Opus 5.5 | $0.00006 | $0.00162 |
| Sonnet 5.5 | $0.00003 | $0.00081 |
| Haiku 4.5 | $0.00002 | $0.00040 |
Grade A, and why
Session Management 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 1mo 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.
What it actually says
Session Management
Use session management guidance when resuming, inspecting, or reasoning about a conversation's persisted runtime state.
Operating Rules
- Resume with the
session_idfrom a previous run. The runtime rebuilds the session from durable state when persistence is enabled. - Treat event logs, catalogs, and projection files as derived. A current
Sessionis domain-only conversation and metadata; the store issues the physical authority used to commit or recover it. Never repair state with ad hoc file edits. - An archived session is lifecycle-terminal and is not an ordinary resume target. Durable-tail recovery can also hold or quarantine a resume when the persisted evidence is ambiguous; preserve the data and surface the typed failure instead of fabricating a repair.
- Compaction may replace old transcript detail with summaries. Use skills, memory, files, and WorkGraph to recover relevant context when needed.
- System messages are ordinary ordered transcript messages. They may repeat at turn boundaries; do not assume there is one mutable system-prompt slot.
- Host-injected context is a typed, separate user-channel message delivered immediately before the turn input. It is excluded from memory indexing and is not a free-form transcript role.
- Session persistence preserves conversation continuity. It does not replace WorkGraph for shared durable work or Schedule for future wakeups.
Compaction
- Lower thresholds compact more often and reduce live context.
- Higher thresholds preserve more live context but cost more tokens.
- Compaction summaries are typed runtime-minted messages. Hosts and transcript rewrite callers cannot mint that role, and summaries are not re-indexed into memory.
- If compaction causes ambiguity, inspect durable artifacts or ask for the missing detail instead of inventing it.
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.
- 1mo ago First seen · 43 lines · 16 tokens per session scan A 96d10afb7975
Session Management is a skill published in the GitHub repository lukacf/meerkat (23 stars, last pushed today), licensed Apache-2.0. It adds 16 tokens to every session and 405 once invoked, about $0.0001 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 2026-08-30.
Other skills, from other repositories
proactive-agent
Transform AI agents from task-followers into proactive partners that anticipate needs and continuously improve. Now with WAL Protocol, Working Buffer, Autonomous Crons, and battle-tested patterns. Part of the Hal Stack 🦞.
memory-recall
Persistent cross-session memory system with fact extraction, semantic search, and compiled memory layers. Your agent remembers everything across sessions — user preferences, project context, past decisions, personal details. Use when: (1) User references something from a previous conversation, (2) Agent needs project…
session-checkpoint
Re-orient a long, dense, or interruption-prone agent session without replacing the underlying evidence. Use when the user asks for a recap, orientation, simple teaching explanation, or spoken summary; when a session has accumulated many turns or tool calls; before context compaction; or when a Brief-Spec hook says a…
memory-continuity
Preserves cross-session continuity with deterministic MEMORY.md, index.md, and snapshot handoffs under .agentic/memories/.
plan-memory-index
Maintains concise indexed memory entries for applied plans under .agentic/memories/ with clear boundaries between memory and plan sources. Memory retrieval is prioritized before loading historical plan files.
llm-wiki
Karpathy's LLM Wiki: build/query interlinked markdown KB.