Fixes F7-F11 from the 2.0.0 script review:
- UTF-8 stdout/stderr in every script; UTF-8 decoding of git/rg/yt-dlp
output and of hook stdin
- utf-8-sig reads, so BOM files keep their frontmatter
- fingerprints match LF or CRLF checkouts (lint, lint_quick.sh,
consolidate); prefix match on the full digest; short ones reported
- POSIX paths in lint output; rg --path-separator /
- hook installers default to the running interpreter, quote paths,
--check reports a missing interpreter, odd settings are an error
New tests/unit/test_windows.py.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fixes F6, F12-F16 from the 2.0.0 script review:
- share: plain file/mailbox names and mailbox-bound share_id/version, or
the message is ignored (no ckb.yaml injection, no writes outside
raw/inbox/); ledger rejects whitespace/control characters; every
commit on a message must be by an owner; duplicate members.yaml
entries unown a mailbox; only verified retractions hide mail; branch,
url and name validated before git; received copy always untrusted;
<private> stripped from title/tldr and unclosed tags; no send without
the secret scan
- capture: frontmatter values quoted unless plainly safe; one-line title
- lint check 13: chat-attributed Crux needs no Sources
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- F11 new opt-in module `capture` (skill ckb-capture, capture_url.py):
stdlib fetch with readable-text extraction, Wayback only over a failed
or thinner live page, --from-file for the agent's own fetch, PDFs saved
as files, --media via yt-dlp with consent, untrusted: true marking,
dedup at capture time; dedup_inbox.py now compares source_url (D-0017)
- F22 tests/run_evals.py (list/prepare/check/validate) and five
structural eval cases; reference run for ckb-decide/record
- F18 manual section with untested setups for Codex CLI, Cursor,
OpenCode and Gemini CLI, plus the Claude Code-only feature table;
AGENTS.md tells agents without skill loading to read SKILL.md
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>