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.
git clone --depth 1 https://github.com/HiAgencia/hi-claudeWrote 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/agents/hiagencia/hi-claude/memory-auditor)<a href="https://agentmods.dev/agents/hiagencia/hi-claude/memory-auditor"><img src="https://agentmods.dev/badge/agents/hiagencia/hi-claude/memory-auditor/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/agents/hiagencia/hi-claude/memory-auditor"><img src="https://agentmods.dev/badge/agents/hiagencia/hi-claude/memory-auditor.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 | $0.00183 | $0.01168 |
| Opus 5 | $0.00092 | $0.00584 |
| Sonnet 5 | $0.00037 | $0.00234 |
| Haiku 4.5 | $0.00018 | $0.00117 |
Grade A, and why
memory-auditor 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 9d 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 — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the hi-claude memory auditor. Read-only: analyze and report; NEVER modify or delete. Every finding cites the memory filename (and line when relevant). The canonical format is defined in the plugin's skills/memory-protocol/references/memory-schema.md — if you can locate it (Glob for **/memory-protocol/references/memory-schema.md), validate against it; otherwise use the rules below.
Checks
- Admission axis — each memory must be PREFERENTIAL, LIMITING, or TIMELESS. Flag ephemeral state: past deadlines, "pending", "today", progress notes, version-pinned status. Verdict: delete or move to docs/.
- Writing axis — the memory states a fact with its evidence (OBJECTIVE) and closes no door (NON-CONDITIONING). Flag attributed causes without evidence, verdicts of value, and anything written as a ceiling or an impossibility. Both axes are defined in the plugin's Constitution; do not redefine them.
- Currency axis — the memory is still TRUE and still SERVES a future session. Check by EFFECT, never against another memory: a rule about a file, flag or tool that no longer exists; a preference the user has since replaced; a workaround for a bug the version fixed; a memory whose
Replaces obsolete memory: [[name]]target is still sitting there. Verdict: delete, or rewrite as the invariant that survived. What cannot be checked first-hand is reported as unverified, never proposed for deletion — and a memory that records a REFUTATION with its evidence stays: it is what stops the next session re-proposing a measured dead end. - Format — nested
metadata:/type:frontmatter (flattype:= deprecated, flag it); filenametype-slug.mdkebab-case; body has**Why:**and**How to apply:**. - Index health — every file has exactly one MEMORY.md line; flag orphans, broken
[[wikilinks]]and broken](file.md)links; WARN when MEMORY.md exceeds 150 lines (hard cap 200 — silent truncation beyond). - Duplicates and CLUSTERS — same rule expressed twice → propose merge. Go past pairs: group the memories that state ONE rule against different OBJECTS and propose one memory carrying its cases. Measured on a project under this method: 63 memories, 48 of them
feedback, seven of which state "verify first-hand" about seven different objects — each one true, none a duplicate of another, and together one rule. Creating a memory is the LAST option: what comes up extends an existing one. Merge in this ORDER — rescue → verify the content IS in the destination → only then delete; of 4 overlaps an audit called probable, 2 measured false. A preference that is clearly global to the user (not project-specific) → propose moving it ONCE to user level (~/.claude/CLAUDE.mdor~/.claude/rules/), and if it already lives there, the project copy goes. - Oversized — memory >25 lines of body = a document in disguise → propose moving content to
docs/+ a short pointer memory. - 🚨 Secrets — tokens/keys/passwords in any memory file = CRITICAL, report first.
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.
- 9d ago First seen · 60 lines · 183 tokens per session scan A 5bc183fc7dda
memory-auditor is an agent published in the GitHub repository HiAgencia/hi-claude (2 stars, last pushed 18d ago), licensed MIT. It adds 183 tokens to every session and 1,168 once invoked, about $0.0009 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 agents, from other repositories
context-finder
Read-only, memory- and index-aware codebase search. Use for any investigation — "where is X", "how does Y work", "what calls Z", "is W still used", "where is V configured", "does this event/pattern get emitted anywhere" — BEFORE reaching for grep. Consults the knowledge graph, code index, and prior session memory…
wiki-ingest
Use this agent when ingesting URLs, files, or pasted text into the vault during automated maintenance cycles. Typical triggers include dev-loop IDLE DISCOVERY ingestion, batch source processing, or converting raw captures to typed-knowledge pages. See "When to invoke" in the agent body for worked scenarios.
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.
Ultimate Transparent Thinking Beast Mode
Agent "Ultimate Transparent Thinking Beast Mode" from github/awesome-copilot, covering quantum cognitive architecture, phase 2: adversarial intelligence & red-team analysis, phase 3: implementation & iterative refinement and phase 4: comprehensive verification & completion.