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 jscott3201/aionforge-memory --skill memory-loopgit clone --depth 1 https://github.com/jscott3201/aionforge-memoryWrote 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/jscott3201/aionforge-memory/memory-loop)<a href="https://agentmods.dev/skills/jscott3201/aionforge-memory/memory-loop"><img src="https://agentmods.dev/badge/skills/jscott3201/aionforge-memory/memory-loop/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/jscott3201/aionforge-memory/memory-loop"><img src="https://agentmods.dev/badge/skills/jscott3201/aionforge-memory/memory-loop.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 | $0.00047 | $0.00704 |
| Opus 5.5 | $0.00019 | $0.00282 |
| Sonnet 5.5 | $0.00009 | $0.00141 |
| Haiku 4.5 | $0.00005 | $0.00070 |
Grade A, and why
memory-loop 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.
How it starts
The opening of the file, as written. The whole thing — 37 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Memory Loop
Requires an enabled Aionforge Memory MCP server.
Use this skill to make memory part of the task loop, not a final afterthought.
Procedure
- Start with
memory-recall. Search broadly enough to find prior decisions, preferences, blockers, release state, and failed attempts. Also recall open work withwork_query/work_treeso you continue the backlog instead of re-deriving it. On every read path —search,read_memory,session_manifest,work_query— assert the teams you belong to (e.g.teams: ["aionforge-memory-team"]); read authorization is per-call, so team memory and team work items are out of scope unless you assert the team on that call. A by-idread_memoryof a team memory needs that team asserted in the same call (parity with search) — it never auto-widens. - Work from current evidence. Recalled memory can guide attention, but repo state, tool output, and user instructions win.
- Capture along the way the moment a durable fact lands: decision made, blocker found, fix verified, release changed, user preference learned, or approach rejected. Do not save these for the end — a context compaction can discard them first.
- Track the work as it moves. When a task, blocker, or TODO appears,
work_createa work item (see thework-trackingskill);work_advanceits status as it progresses. Tasks are work items, not memory episodes — and there is no "note" to store directly. - Be generous with memory. Aionforge can handle large memory sets; several precise records are better than one vague end note.
- At natural checkpoints, search again if new terms, file paths, ids, or failures appear.
- Before ending, capture a handoff when future agents would benefit: branch, PR, commits, tests, CI, remaining work, and caveats. Leave the remaining work as work items so the next agent can
work_queryit. When the handoff is directed at a specific agent or teammate — not just left for whoever comes next — alsomessage_sendthem a brief so they actually receive it (see theagent-messagingskill): the capture and work items make it durable, the message delivers it. - Run
consolidation_status; runconsolidateonly when the approval policy permits mutating derived memory.
What ships with it
1 file 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.
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 · 37 lines · 47 tokens per session scan A f799a430fce1
memory-loop is a skill published in the GitHub repository jscott3201/aionforge-memory (11 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 47 tokens to every session and 704 once invoked, about $0.0002 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 2026-08-31.
Other skills, from other repositories
brain-write
How to write to the INITE Brain knowledge graph from an agent loop — recordfact (with the conversationId / evidence[] grounding inputs), ingestdocument, linkentities, retractfact, recordfeedback, and the detectcontradiction preflight. Covers confidence picking, claim grounding, retract vs forget semantics, identityof…
brain-conflict
How to detect and resolve conflicting beliefs in the INITE Brain knowledge graph — the COMPETING fact status, getcompetingfacts, detectcontradiction preflight, and the human-in-the-loop adjudication workflow. Use when the timeline shows two facts disagreeing on the same predicate, or when an agent needs to decide what…
brain-recall
Recall everything brain knows about one specific entity — current profile, full bitemporal timeline, graph neighbours, and unresolved disagreements. Use when the user names a person/company/thing and asks "tell me about them", "what's their history", or "what do we still disagree about?". For a single-shot LLM…
brain-bitemporal
How to query the INITE Brain knowledge graph across time — the asOf parameter, validFrom/validUntil semantics, reading retracted facts, and the memorydiff "what changed between two cursors" surface. Use when the user's question has a temporal dimension ("on X date", "before Y", "what's new since last conversation").
memory-review
Memory hygiene audit -- finds stale facts, contradictions, low-confidence entries, and consolidation candidates in Pensyve memory. Use periodically to maintain memory quality.
session-memory
End-of-session memory capture -- classifies session signals using a tiered taxonomy and stores confirmed items via Pensyve. Use when ending a work session or when the user wants to capture what was learned.