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 a-Fig/Accordion --skill accordion-context-recallgit clone --depth 1 https://github.com/a-Fig/AccordionWrote 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/a-fig/accordion/accordion-context-recall)<a href="https://agentmods.dev/skills/a-fig/accordion/accordion-context-recall"><img src="https://agentmods.dev/badge/skills/a-fig/accordion/accordion-context-recall/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/a-fig/accordion/accordion-context-recall"><img src="https://agentmods.dev/badge/skills/a-fig/accordion/accordion-context-recall.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 | $0.00089 | $0.00503 |
| Opus 5 | $0.00044 | $0.00251 |
| Sonnet 5 | $0.00018 | $0.00101 |
| Haiku 4.5 | $0.00009 | $0.00050 |
Grade A, and why
accordion-context-recall 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.
What it actually says
Accordion, an external desktop app, may be folding parts of your context to save tokens. A folded block appears as:
{#3f9a2c FOLDED} Assistant analyzed the test failures: three imports were missing…
The 3f9a2c is a fold code — the short summary after FOLDED} is a digest. The original content is preserved and always retrievable.
recall — read folded content this turn
Call recall with one or more codes copied verbatim from the markers:
recall({codes: ["3f9a2c"]})
recall({codes: ["3f9a2c", "00a2cd"]})
The full original content of each block comes back as this tool's result, immediately — like reading a file. Your standing context is not changed: the block stays folded, costs no extra tokens, and the human's view is untouched.
This is the right choice when:
- You need a specific value, piece of code, or detail for the current step only
- You don't want the block to permanently reopen (it stays folded for future turns)
- You're under a tight token budget and opening it permanently would trigger more folding
When to use unfold instead
unfold({codes: [...]}) forces the block open in your standing context — the full content appears on your next turn and stays open until the human re-folds it. Use unfold when you will keep referencing the block over several turns, not just once.
Summary:
- recall — read once, now. No context change. Content in the tool result this turn.
- unfold — keep it open. Content in your context next turn, stays open.
Prefer recall for one-off reads; use unfold only when repeated access justifies the token cost of keeping it open.
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 · 39 lines · 89 tokens per session scan A 363de83c9b06
accordion-context-recall is a skill published in the GitHub repository a-Fig/Accordion (243 stars, last pushed 19d ago), licensed MIT. It adds 89 tokens to every session and 503 once invoked, about $0.0004 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 2026-08-30.
Other skills, from other repositories
codebase-memory
Use the codebase knowledge graph for structural code queries. Triggers on: explore the codebase, understand the architecture, what functions exist, show me the structure, who calls this function, what does X call, trace the call chain, find callers of, show dependencies, impact analysis, dead code, unused functions…
memory-compounding
Review and sharpen persistent memory so it compounds instead of accumulating. Use when pruning pi-hermes-memory entries, doing monthly memory hygiene, or when the same lesson has been recorded multiple times.
session-handoff
Hand off work to another Pi session — write a handoff doc, save key decisions to memory, and notify the target session via intercom. Use when a session is getting long, when you need to continue work in a fresh context, or when transferring work between projects/sessions. Composes handoff.ts (doc writer) + memory…
aha
Capture aha-moments as durable insight notes under .ai/insight/, or quiz the user on past insights to reinforce recall. Use when the user has just learnt something non-obvious (a fix they didn't know, an explanation that clicked, a gotcha discovered) and wants to record it, or when the user invokes /aha quiz to be…
persist
Store your findings, analysis, or other outcomes on disk for future reference. Use when user asks you to do something that does not explicitly mention planning but clearly does not involve taking action yet e.g. analysing or investigating something.
session-search
Search and verify Joel's Pi, Claude, Codex, Cursor, Grok, and OpenCode session history. Use when asked to recover prior context, find exact transcript evidence, inspect session receipts, continue old work, or verify capture health. Prefer flowing recall before raw transcript search when the question is about prior…