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/waittim/MemoryCustodianWrote 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/commands/waittim/memorycustodian/memory-compact)<a href="https://agentmods.dev/commands/waittim/memorycustodian/memory-compact"><img src="https://agentmods.dev/badge/commands/waittim/memorycustodian/memory-compact/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/commands/waittim/memorycustodian/memory-compact"><img src="https://agentmods.dev/badge/commands/waittim/memorycustodian/memory-compact.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 | $0.00000 | $0.00329 |
| Opus 5.5 | $0.00000 | $0.00132 |
| Sonnet 5.5 | $0.00000 | $0.00066 |
| Haiku 4.5 | $0.00000 | $0.00033 |
Grade A, and why
memory-compact 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 21d 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
memory-compact
Run a dry run first:
memory-custodian compact
The report lists exact duplicate top-level bullet units, exact tombstone matches, and candidates requiring Agent review. Each top-level bullet unit includes its continuation and nested lines; nested bullets are never cleanup candidates on their own. For each candidate, determine scope, type, confidence, and whether equivalent memory already exists. Then edit the appropriate Markdown directly or use memory-custodian add.
If the exact mechanical cleanup is appropriate, run:
memory-custodian compact --apply --confirm-plan <PLAN_ID>
This command does not classify or promote candidates. It only removes the exact duplicate complete units and tombstone matches shown in the preview; reviewed candidates remain in the inbox until handled explicitly. Protocol 0.8 applies the multi-file write through the shared transaction journal; if it is interrupted, run audit --transactions and recover by opaque transaction ID. Run memory-custodian check after semantic updates.
For an over-budget active file, run a target dry run:
memory-custodian compact --target decisions.md
First shorten decisions over 120 tokens, merge superseded entries, move subsystem knowledge into matched areas, and retain active invariants in normal loading paths. Then apply only after reviewing the plan:
memory-custodian compact --target decisions.md --apply --archive-oldest --confirm-plan <PLAN_ID>
For semantic compaction, read skills/memory-custodian/references/compaction-policy.md and update memory files directly.
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.
- 21d ago Changed e3f06fbf790a
- 1mo ago First seen · 32 lines · 0 tokens per session scan A 475b2ef45ee8
memory-compact is a command published in the GitHub repository waittim/MemoryCustodian (22 stars, last pushed 20d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 329 tokens. 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 commands, from other repositories
ctx-drift
Find contradictions, gaps, and inconsistencies between AI context files from different tool ecosystems.
ctx-check
Validate claims in AI context files against the actual codebase to find stale, broken, and drifted references.
ctx-fix
Fix staleness and drift issues found by /ctx:check and /ctx:drift with interactive proposed edits.
rem_help
Each section is printed verbatim by --help for the command it names. The agreed design is issue #1656; change a page there and here together.
speckit.archive
Archive a feature specification into main project memory after merge, resolving gaps and conflicts.
memory-query
Query an existing graphify graph before rereading the full corpus, using query, path, or explain flows.