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/sefaertunc/WorclaudeWrote 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/sefaertunc/worclaude/update-claude-md)<a href="https://agentmods.dev/commands/sefaertunc/worclaude/update-claude-md"><img src="https://agentmods.dev/badge/commands/sefaertunc/worclaude/update-claude-md/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/sefaertunc/worclaude/update-claude-md"><img src="https://agentmods.dev/badge/commands/sefaertunc/worclaude/update-claude-md.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 | $0.00015 | $0.00988 |
| Opus 5 | $0.00008 | $0.00494 |
| Sonnet 5 | $0.00003 | $0.00198 |
| Haiku 4.5 | $0.00002 | $0.00099 |
Grade A, and why
update-claude-md 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 8d 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 — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Propose updates to CLAUDE.md based on this session's work AND the project's captured learnings. Apply changes only with explicit per-change consent.
Sources of proposals
Look at three places, in priority order:
-
.claude/learnings/directory — surface promotion candidates.learn-capture.cjswrites one file per[LEARN]category and appends when the same category is captured again. The directory'sindex.jsonupdatescreatedto the latest capture date — treat it as "last touched," not "first created." See thememory-architectureskill for the full layer model.Concrete promotion algorithm — a learning is a candidate when at least one of these holds:
- Recurrence: the file at
.claude/learnings/<category>.mdcontains 3 or more**Rule:**blocks. Count by scanning the file directly (the index doesn't track count). Three independent captures of the same category is the threshold for "this isn't a fluke." - Recency: the index entry's
createddate is within the last 14 days. Recent learnings are warmer signals than old ones. - Drift: the learning's content is structurally relevant to an
existing
CLAUDE.mdsection (e.g., a "always do X" pattern that belongs in## Critical Rulesor## Gotchas) but the rule is missing fromCLAUDE.md.
A candidate satisfying only recency is weak — surface it but rank it below a recurring or drift-aligned candidate. Prefer learnings that hit two or three signals at once.
Out of scope for now: Claude Code's auto memory at
~/.claude/projects/<slug>/memory/is NOT scanned. Auto memory is personal-scoped; promotion toCLAUDE.md(team-scoped) requires the user's explicit reasoning and is mediated through/learn. This may be revisited once Phase 6a observability ships. - Recurrence: the file at
-
This session's mistakes — if Claude made the same mistake twice in this session, it's a candidate for a Gotchas entry. Once-per-session mistakes are not yet rule-worthy.
-
This session's discoveries — non-obvious patterns the user confirmed are worth documenting (e.g., "we always do X here because Y").
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.
- 8d ago First seen · 108 lines · 15 tokens per session scan A fe09780b437b
update-claude-md is a command published in the GitHub repository sefaertunc/Worclaude (4 stars, last pushed 29d ago), licensed MIT. It adds 15 tokens to every session and 988 once invoked, about $0.0001 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 commands, from other repositories
context-stats
Display context window usage and token statistics.
handoff
Create a handoff document for seamless session continuity.
learn
Force-extract a lesson from the last error, bypassing the 2-occurrence confidence gate.
lessons
Browse, search, and manage learned lessons. Usage: /lessons [search|stats|prune|export].
forget
Remove or archive a specific lesson. Usage: /forget.
export
Export instincts for sharing.