The post-edit-format hook was hardcoded to use Prettier. Projects using
Biome had their code reformatted with Prettier defaults (e.g. double
quotes overwriting single quotes).
Now the hook walks up from the edited file to find the project root,
then checks for config files:
- biome.json / biome.jsonc → runs Biome
- .prettierrc / prettier.config.* → runs Prettier
- Neither found → skips formatting silently
New community skills: content-hash-cache-pattern, cost-aware-llm-pipeline,
regex-vs-llm-structured-text, swift-actor-persistence, swift-protocol-di-testing
The PreToolUse Write hook blocks creation of .md files to prevent
unnecessary documentation sprawl. However, it also blocks Claude Code's
built-in plan mode from writing plan files to .claude/plans/*.md,
causing "BLOCKED: Unnecessary documentation file creation" errors
every time a plan is created or updated.
Add .claude/plans/ path to the whitelist so plan files are not blocked.
Comprehensive coding standards for modern C++ (C++17/20/23) derived
from isocpp.github.io/CppCoreGuidelines. Covers philosophy, functions,
classes, resource management, expressions, error handling, immutability,
concurrency, templates, standard library, enumerations, naming, and
performance with DO/DON'T examples and a pre-commit checklist.
The inline environment variable syntax `TIMESTAMP="$timestamp" echo ...`
does not work correctly because:
1. The pipe creates a subshell that doesn't inherit the variable
2. The environment variable is set for echo, not for the piped python
Fixed by using `export` and separating the commands:
- export TIMESTAMP="$timestamp"
- echo "$INPUT_JSON" | python3 -c "..."
This ensures the TIMESTAMP variable is available to the python subprocess.
Fixes#227
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Round 125: Tests for readFile returning garbled strings (not null) on binary
files, output() handling undefined/NaN/Infinity as non-objects logged directly
(and JSON.stringify converting NaN/Infinity to null in objects), and loadAliases
with __proto__ key in JSON proving no prototype pollution occurs.
Total: 935 tests, all passing.
Round 124: Tests for findFiles matching dotfiles (unlike shell glob where *
excludes hidden files), getAllSessions strict date equality filter (wrong format
silently returns empty), and parseSessionMetadata title regex edge cases
(no space after #, ## heading, multiple H1, greedy \s+ crossing newlines).
Total: 932 tests, all passing.
Round 123: Tests for countInFile non-overlapping regex match behavior (aaa with
/aa/g returns 1 not 2), replaceInFile with $& and $$ substitution tokens in
replacement strings, and parseSessionMetadata CRLF section boundary bleed where
\n\n fails to match \r\n\r\n. Total: 929 tests, all passing.
Round 122: Tests for findFiles glob dot escaping (*.txt must not match filetxt),
listAliases limit=0/negative/NaN returning all due to JS falsy check, and
getSessionById matching old YYYY-MM-DD-session.tmp filenames via noIdMatch path.
Total: 926 tests, all passing.
- parseSessionFilename rejects Feb 29 in century non-leap years (1900, 2100) but accepts 2000/2400
- replaceInFile with /(?:)/g zero-width regex inserts at every position boundary
- parseSessionMetadata preserves raw markdown formatting (**bold**, `code`, _italic_) in titles
Total: 899 tests
Round 110: Tests for findFiles with unreadable root directory returning
empty array (vs Round 71 which tested subdirectory), parseSessionFilename
year 0000 exposing JS Date 0-99→1900-1999 mapping quirk, and uppercase
session ID rejection by [a-z0-9]{8,} regex. 890 tests total.
Round 109:
- appendFile creating new file in non-existent directory (ensureDir + appendFileSync)
- getExecCommand with ../ path traversal in binary (SAFE_NAME_REGEX allows ../)
- getAllSessions skips .tmp files that don't match session filename format
- grepFile with ^$ pattern verifies empty line matching including trailing newline phantom
- replaceInFile with self-reintroducing replacement confirms single-pass behavior
- setAlias with whitespace-only title exposes missing trim validation vs sessionPath
- countInFile with named capture groups verifies match(g) ignores group details
- grepFile with multiline (m) flag confirms flag is preserved unlike stripped g
- getAllSessions with array/object limit tests Number() coercion edge cases
- parseSessionMetadata: blank line within Completed section truncates items
due to regex lookahead (?=###|\n\n|$) stopping at \n\n boundary
- grepFile: sticky (y) flag not stripped like g flag, causing stateful
.test() behavior that misses matching lines
- getExecCommand: object args bypass SAFE_ARGS_REGEX (typeof !== 'string')
but coerce to "[object Object]" in command string
