- Run cargo fmt on diff_cmd.rs and init.rs (line length violations)
- Add Unreleased section to CHANGELOG.md with the two bug fixes
from this PR (required by doc review CI gate)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
Keep truncation accuracy fix (HEAD) — accurate overflow counts in
condense_unified_diff. Drop test_condense_unified_no_truncation from
develop (incompatible). Add test_no_truncation_large_diff and
test_long_lines_not_truncated from develop (test compute_diff, no conflict).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two silent truncation bugs in git.rs:
1. compact_diff: hunk truncation showed "... (truncated)" with no count.
An LLM had no way to know if 5 or 450 lines were dropped. Replace
with "... (N lines truncated)" by tracking hunk_skipped separately
from hunk_shown — flush the count at each hunk/file boundary.
2. filter_log_output: commit body beyond 3 lines was silently dropped.
A 20-line body explaining a breaking change would show 3 lines with
no indicator that 17 were omitted. Now appends "[+N lines omitted]"
when body_omitted > 0.
Adds 2 RED-then-GREEN tests to lock in the new behavior.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
The overflow message "+N more" in condense_unified_diff was lying.
The `changes` vec was capped at 15 entries, so `changes.len() - 10`
could only reach 5 — reporting "+5 more" when 190 lines were actually
truncated. Fix: track `added + removed` directly and compute true
overflow as `(added + removed) - 10`.
Adds 7 accuracy tests across 4 modules to lock in correct overflow
reporting:
- diff_cmd: overflow count matches true total (200 changes → "+190 more")
- diff_cmd: no spurious overflow message for 8 changes
- git: format_status overflow count is exact (25 staged → "+10 more")
- git: compact_diff recovery hint present when hunk is truncated
- grep: documents uncapped-vector invariant that prevents the diff bug
- filter: smart_truncate kept + reported_more == total_lines
Closes the test gap flagged in testing-patterns.md backlog for diff_cmd.rs.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
Changed default filter level from "minimal" to "none". RTK read now
shows complete file content by default. Filtering is opt-in:
rtk read file.rs # full content (was: minimal filter)
rtk read file.rs -l minimal # light filtering (opt-in)
rtk read file.rs -l aggressive # signatures only (opt-in)
Also adds fallback: if a filter produces empty output on non-empty
file, show raw content with a warning.
Fixes#822
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
- Detect binary files (null bytes in first 8KB) before filtering
- Show clear message: [binary file] path (size) instead of empty output
- Fallback to raw content if filter produces empty output on non-empty file
- Prevents LLM from concluding a 70MB file is "empty" (was #822)
Fixes#822
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
Silent truncation caused a user to make irreversible decisions on
incomplete data. rtk diff was capping at 50 changes and 70-80 char
lines without clear warning to the LLM.
- Remove all line truncation (70/80 char limits)
- Remove change count limit (was 50, now unlimited)
- Remove truncate import (no longer needed)
- Same fix for condense_unified_diff (stdin/git diff path)
- Add 3 tests verifying zero truncation on large inputs
The only compression remaining is the summary header
(+N added, -N removed, ~N modified) and stripping diff metadata.
Fixes#827
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
- Add `rtk init --copilot` that creates .github/hooks/rtk-rewrite.json
and .github/copilot-instructions.md in the current project
- Hook routes through `rtk hook copilot` (auto-detects VS Code vs CLI)
- VS Code Copilot Chat: transparent rewrite via updatedInput
- Copilot CLI: deny-with-suggestion (CLI limitation)
- Update README: split Copilot VS Code/CLI, add Vibe status, 10 tools
Closes#823
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
Commands like `git status 2>&1` or `cargo test 2>/dev/null` were not
rewritten because the redirect suffix prevented pattern matching.
Now strips redirects (2>&1, 2>/dev/null, etc.) before matching,
then re-appends them to the rewritten command.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
- Add "Privacy & Telemetry" section to README explaining what is
collected, what is NOT, and how to opt-out
- Show telemetry notice during rtk init so users are informed
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
When `gh pr diff` is called with output-format-changing flags like
--name-only, --stat, --name-status, --numstat, or --shortstat, the
output is a plain filename/stat list rather than a unified diff.
compact_diff() expects diff headers and hunks, so it produces empty
output from these formats.
Skip filtering and passthrough directly when any of these flags are
present. The output is already compact and doesn't benefit from diff
compaction.
Fixes#730
Signed-off-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Add --comments to the passthrough trigger list in should_passthrough_pr_view().
The --comments flag changes gh output to include comments, but the rtk
filter hardcodes a JSON field list without 'comments', silently
discarding all comment data. This misleads AI agents into concluding
issues have no responses.
Add should_passthrough_issue_view() for view_issue(), which previously
had no passthrough check at all. Both functions now passthrough on
--comments, --json, --jq, and --web.
Fixes#720
Signed-off-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Add crate::tee::tee_and_hint call to playwright_cmd::run(), matching
the pattern used by all other test runners (vitest, pytest, cargo,
go, rspec, rake). When playwright fails, the full raw output is saved
to ~/.local/share/rtk/tee/ and a one-line hint is appended so the
LLM can read the complete output without re-running.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Ousama Ben Younes <benyounes.ousama@gmail.com>
format_compact() was truncating error messages to 2 lines via .take(2).
Playwright (and vitest, pytest, cargo test) errors contain the critical
expected/received diff and call log starting at line 3+, so agents saw
only the error type with no actionable debug information.
Fix: iterate all lines of error_message for each failure instead of
taking the first 2. Summary line stays compact; only the per-failure
detail is preserved in full.
Affects all test runners using TestResult (playwright, vitest, cargo,
pytest, dotnet) — any test failure is now fully visible.
Closes#690
Signed-off-by: Ousama Ben Younes <benyounes.ousama@gmail.com>
Bug 2 (Critical): *:* catch-all matched nothing. strip_suffix('*') left
"*:" which after trim became "*" (non-empty), so the branch returned
false instead of true. Fix: detect empty-or-star prefix after stripping.
Bug 3 (Medium): leading wildcards ("* --force"), middle wildcards
("git * main"), and multi-wildcard patterns ("git * --force *") fell
through to exact match, silently failing. Add glob_matches() with
character-level segment anchoring: first segment must be prefix, last
must be suffix, middle segments found via str::find in order.
Colon normalization in glob_matches(): "sudo:*" -> "sudo *" so both
fast path and glob path interpret colon syntax consistently.
New tests: test_star_colon_star_matches_everything,
test_leading_wildcard, test_leading_wildcard_no_partial,
test_middle_wildcard, test_middle_wildcard_no_match,
test_multiple_wildcards, test_deny_with_leading_wildcard,
test_deny_star_colon_star.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
Bug 1 (Critical): check_command() was called inside Some(rewritten),
so non-RTK commands (rm, kill, python3 -c) bypassed deny rules entirely.
Move verdict check before registry::rewrite_command() so all commands
are evaluated regardless of whether RTK has an equivalent.
Bug 4 (Medium): print!() before process::exit() could leave stdout
unflushed. Add explicit std::io::stdout().flush() after each print!().
Add Eq derive to PermissionVerdict (required for == comparison).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
The PreToolUse hook was emitting `permissionDecision: "allow"` on every
rewritten command, bypassing deny and ask rules in .claude/settings.json.
- Add `src/permissions.rs`: loads Bash deny/ask rules from all 4 Claude
Code settings files (project + global, settings.json + settings.local.json),
checks commands (including compound && / || / | / ;) and returns
Allow / Deny / Ask verdict. 16 unit tests.
- Modify `src/rewrite_cmd.rs`: after finding a rewrite, check the original
command against permissions. Exit 0 = allow (auto-approve rewrite),
exit 2 = deny (passthrough, let CC native deny handle it),
exit 3 = ask (print rewrite but no permissionDecision, CC prompts user).
- Update both hook files to handle exit codes 2 and 3. Version bumped 2→3.
- Bump `CURRENT_HOOK_VERSION` 2→3 in `hook_check.rs` so users with the old
hook get the upgrade prompt.
- Fix set -euo pipefail bug in .claude/hooks/rtk-rewrite.sh: capture exit
code with `|| EXIT_CODE=$?` instead of bare assignment.
Fixes#260
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
rake test ignores positional file arguments and only supports TEST=path
for single-file runs. When users pass positional test files (e.g.,
`rtk rake test file1.rb file2.rb` or `rtk rake test file.rb:15`),
select_runner() now switches to `rails test` which handles single files,
multiple files, and line-number syntax natively.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Nicholas Lee <nicholas.lee@appfolio.com>
- git diff: raise max_hunk_lines from 30 to 100 (LLMs need full hunks)
- git log: show 3 body lines instead of 1 (preserves BREAKING CHANGE, migration notes)
- json: show values by default (LLMs need values for config debugging), add --schema for types-only
Tested with phi4:14b on local LLM — all 3 fixes improve comprehension.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
golangci-lint v2 removed --out-format=json in favour of --output.json.path stdout.
Detect the installed major version at runtime via golangci-lint --version and branch
on the correct flag and JSON extraction strategy.
- Use --output.json.path stdout for v2, --out-format=json for v1
- Extract JSON from first line only on v2 (v2 appends trailing metadata after JSON line)
- Deserialise new v2 fields: SourceLines, Severity, Offset (serde default for v1 compat)
- Show first source line per linter-file group on v2 for richer context
- Always forward stderr to caller (was silently dropped unless --verbose)
- Falls back to v1 behaviour on any version detection failure
Signed-off-by: Adam Powis <a.powis.91@gmail.com>
Install RTK rules in .clinerules (project-scoped) so Cline
prefixes shell commands with rtk for token savings.
Same rules-based approach as Windsurf and Codex.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
Install RTK rules in .windsurfrules (project-scoped) so Cascade
prefixes shell commands with rtk for token savings.
Windsurf hooks don't support command rewriting (only blocking),
so RTK uses the rules-based approach (like Codex with AGENTS.md).
Tested: Windsurf Cascade correctly uses rtk git status after install.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
Add `rtk hook copilot` command that handles both VS Code Copilot Chat
(updatedInput rewrite) and GitHub Copilot CLI (deny-with-suggestion).
- Auto-detects format: snake_case (VS Code) vs camelCase (Copilot CLI)
- Delegates to `rtk rewrite` (single source of truth)
- 14 hook tests (format detection, rewrite gating, output shape)
- .github/hooks/rtk-rewrite.json for repo-scoped hook config
- .github/copilot-instructions.md for RTK awareness
- Test script: hooks/test-copilot-rtk-rewrite.sh
Rebased on develop (includes Gemini #573, Codex #377, OpenClaw #358).
Original work by @jeziellopes, cleaned up and rebased by maintainer.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
Co-authored-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
- Add `rtk hook gemini` command: native Rust hook processor for Gemini CLI
BeforeTool hooks. Reads JSON from stdin, delegates to `rewrite_command()`
(single source of truth), outputs Gemini-format JSON response.
- Add `--gemini` flag to `rtk init`: installs hook wrapper script, GEMINI.md,
and patches ~/.gemini/settings.json with BeforeTool hook entry.
- Add `rtk init -g --gemini --uninstall`: clean removal of all Gemini artifacts.
- 6 unit tests covering hook format, rewrite delegation, and exclusions.
Replaces PR #174 which had too many conflicts after upstream restructuring.
Signed-off-by: Ousama Ben Younes <benyounes.ousama@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* fix: remove decorative emojis from CLI output (#511)
Replace decorative emojis with plain text to reduce token waste.
Keep functional symbols (⚠️ ✓ ❌✅ℹ️) that convey meaning in fewer tokens.
* fix: remove remaining decorative emojis from find_cmd and formatter
Missed in initial emoji cleanup pass: 📁 in find_cmd.rs and parser/formatter.rs
* fix: remove all decorative emojis from CLI output (#511)
Replace emojis with plain text tokens across all production files
for better LLM compatibility. Test fixtures and external tool
detection patterns (e.g. Black's "All done!") are preserved.
* fix: remove last decorative emoji from next_cmd.rs
Remove ⚡ from Next.js Build header, missed in previous passes.
* fix: remove remaining emojis from gh_cmd.rs and init.rs
Replace production emojis:
- gh_cmd.rs: 🟣→[merged], ⚪→[unknown]/[pending], ⭐→removed, 🔱→removed
- init.rs: ⚪→[--] for "not found" status indicators
* fix: remove all checkmark emojis from CLI output
Replace ✓ (U+2713) with plain text across 19 files:
- "ok ✓" → "ok" (git add/commit/push/pull)
- "✓ cargo test: ..." → "cargo test: ..." (all tool summaries)
- Preserved ✓ in input detection patterns and test fixtures
LLMs cannot interpret emoji semantics; plain text is clearer.
---------
Signed-off-by: Patrick Szymkowiak <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
Signed-off-by: Patrick Szymkowiak <patrick@rtk-ai.app>
* fix: remove decorative emojis from CLI output (#511)
Replace decorative emojis with plain text to reduce token waste.
Keep functional symbols (⚠️ ✓ ❌✅ℹ️) that convey meaning in fewer tokens.
Signed-off-by: Patrick Szymkowiak <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: remove remaining decorative emojis from find_cmd and formatter
Missed in initial emoji cleanup pass: 📁 in find_cmd.rs and parser/formatter.rs
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: remove all decorative emojis from CLI output (#511)
Replace emojis with plain text tokens across all production files
for better LLM compatibility. Test fixtures and external tool
detection patterns (e.g. Black's "All done!") are preserved.
Signed-off-by: Patrick Szymkowiak <patrick@rtk-ai.app>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: remove last decorative emoji from next_cmd.rs
Remove ⚡ from Next.js Build header, missed in previous passes.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: remove remaining emojis from gh_cmd.rs and init.rs
Replace production emojis:
- gh_cmd.rs: 🟣→[merged], ⚪→[unknown]/[pending], ⭐→removed, 🔱→removed
- init.rs: ⚪→[--] for "not found" status indicators
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: remove all checkmark emojis from CLI output
Replace ✓ (U+2713) with plain text across 19 files:
- "ok ✓" → "ok" (git add/commit/push/pull)
- "✓ cargo test: ..." → "cargo test: ..." (all tool summaries)
- Preserved ✓ in input detection patterns and test fixtures
LLMs cannot interpret emoji semantics; plain text is clearer.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
---------
Signed-off-by: Patrick Szymkowiak <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
Signed-off-by: Patrick Szymkowiak <patrick@rtk-ai.app>
* fix: P1 exit codes, grep regex perf, SQLite WAL (#631)
* fix: P1 exit codes, grep regex perf, SQLite concurrency
Exit code propagation (same pattern as existing modules):
- wget_cmd: run() and run_stdout() now exit on failure
- container: docker_logs, kubectl_pods/services/logs now check
status before parsing JSON (was showing "No pods found" on error)
- pnpm_cmd: replace bail!() with eprint + process::exit in
run_list and run_install
Performance:
- grep_cmd: compile context regex once before loop instead of
per-line in clean_line() (was N compilations per grep call)
Data integrity:
- tracking: add PRAGMA journal_mode=WAL and busy_timeout=5000
to prevent SQLite corruption with concurrent Claude Code instances
Signed-off-by: Patrick <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: address review findings on P1 fixes
- tracking: WAL pragma non-fatal (NFS/read-only compat)
- wget: forward raw stderr on failure, track raw==raw (no fake savings)
- container: remove stderr shadow in docker_logs, add empty-stderr
guard on all 4 new exit code paths for consistency with prisma pattern
Signed-off-by: Patrick <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
---------
Signed-off-by: Patrick <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: raise output caps for P0 bugs (#617, #618, #620) (#630)
* fix: raise output caps for grep, git status, and parser fallback (#617, #618, #620)
- grep: per-file match cap 10 → 25, global max 50 → 200
- git status: file list caps 5/5/3 → 15/15/10
- parser fallback: truncate 500 → 2000 chars across all modules
These P0 bugs caused LLM retry loops when RTK returned less signal
than the raw command, making RTK worse than not using it.
Fixes#617, #618, #620
Signed-off-by: Patrick <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: update README example and add truncation tests for modified/untracked
- parser/README.md: update example from 500 → 2000 to match code
- git.rs: add test_format_status_modified_truncation (cap 15)
- git.rs: add test_format_status_untracked_truncation (cap 10)
Signed-off-by: Patrick <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* refactor: extract output caps into [limits] config section
Move hardcoded caps into config.toml so users can tune them:
[limits]
grep_max_results = 200 # global grep match limit
grep_max_per_file = 25 # per-file match limit
status_max_files = 15 # staged/modified file list cap
status_max_untracked = 10 # untracked file list cap
passthrough_max_chars = 2000 # parser fallback truncation
All 8 modules now read from config::limits() instead of hardcoded
values. Defaults unchanged from previous commit.
Signed-off-by: Patrick <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
---------
Signed-off-by: Patrick <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* feat(.claude): add /rtk-triage skill — cross-analysis of PRs and issues (#662)
* feat(.claude): add /rtk-triage skill — orchestrated PR+issue cross-analysis
New skill that runs issue-triage + pr-triage in parallel then produces
a cross-analysis layer that neither skill can do individually:
- Double coverage detection: identifies when 2+ PRs target the same issue
(via body scan + file overlap), recommends which to keep/close
- Security gap detection: for security review issues, maps each finding
to a PR (or flags it as uncovered)
- P0/P1 bugs without PR: groups by pattern to suggest sprint batching
- Our dirty PRs: identifies probable cause (conflict with sibling PR,
needs rebase, missing linked issue)
Output is saved automatically to claudedocs/RTK-YYYY-MM-DD.md.
Usage: /rtk-triage (French, auto-save)
/rtk-triage en (English output)
Signed-off-by: Florian Bruniaux <florian@bel-etage.com>
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
* docs(architecture): update module count to 66
Sync ARCHITECTURE.md with current main.rs state.
Previous count (60) was stale since several modules were added
(dotnet_cmd, dotnet_format_report, dotnet_trx, npm_cmd, gt_cmd, etc.).
Signed-off-by: Florian Bruniaux <florian@bel-etage.com>
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
---------
Signed-off-by: Florian Bruniaux <florian@bel-etage.com>
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
* fix: subcommand routing drops unrecognized subcommands (#600) (#601)
- git stash: pass unknown subcommands (save, branch, clear) through
instead of silently falling back to git stash push
- git branch: add --show-current, --set-upstream-to, --format, --sort
to flag detection so they don't get overridden by -a injection
- pip: replace bail!() with passthrough for unknown subcommands
(freeze, download, wheel, etc.)
Fixes#600
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: resolve cargo fmt + 54 clippy warnings blocking CI (#663)
cargo fmt diffs in config.rs, git.rs, playwright_cmd.rs were failing
the fmt CI check, which cascaded to block clippy/test/security on
PRs #632, #635, #638. Also fixes all clippy warnings: dead code
annotations, iterator simplifications, assert patterns, and
unnecessary allocations.
Signed-off-by: Patrick Szymkowiak <patrick@rtk-ai.app>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: discover absolute paths + git global options (#485, #163) (#518)
* fix: discover classifies absolute paths like /usr/bin/grep (#485)
Normalize absolute binary paths before classification:
/usr/bin/grep → grep, /bin/ls → ls, /usr/local/bin/git → git
Adds strip_absolute_path() helper + 5 tests.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: discover and rewrite support git global options -C, --no-pager, etc. (#163)
Strip git global options (-C <path>, -c <key=val>, --git-dir, --work-tree,
--no-pager, --no-optional-locks, --bare, --literal-pathspecs) before
classification so git -C /tmp status is recognized as rtk git.
Rewrite preserves global options: git -C /tmp status → rtk git -C /tmp status
Adds GIT_GLOBAL_OPT lazy_static regex + strip_git_global_opts() helper + 6 tests.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
---------
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: prevent double `--` separator in cargo clippy with -p flags (#519)
When running `rtk cargo clippy -p my-crate -- -D warnings`, Clap with
`trailing_var_arg = true` preserves the `--` in parsed args when flags
precede it. `restore_double_dash()` then added a second `--`, producing
`cargo clippy -p my-crate -- -- -D warnings`. This caused rustc to
interpret `-D` as a filename instead of a lint flag.
Fix: skip restoration when args already contain `--` (Clap preserved it).
Fixes#496
Signed-off-by: Ousama Ben Younes <benyounes.ousama@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* ci: add PR template + target branch check (#521)
- PR template reminds contributors to target develop
- CI workflow labels PRs targeting master with 'wrong-base' and posts a comment
- Excludes develop→master PRs (maintainer releases)
Signed-off-by: Patrick <patrick@rtk-ai.com>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: prevent rtk read from corrupting JSON/YAML/data files (#522)
Add Language::Data variant for data formats (JSON, YAML, TOML, XML, CSV, etc.)
with empty comment patterns to prevent comment stripping. AggressiveFilter
falls back to MinimalFilter for data files.
Fixes#464
Signed-off-by: Ousama Ben Younes <benyounes.ousama@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* fix: skip rewriting find/fd in pipes to preserve xargs compatibility (#439) (#563)
rtk find outputs a grouped format incompatible with pipe consumers
like xargs, grep, wc, sort. Skip rewrite when find/fd is followed
by a pipe, preserving native one-per-line output.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: add hint when git diff is truncated + fix --no-compact passthrough (#427) (#564)
When compact_diff truncates output, append a hint line so Claude knows
how to get the full diff: [full diff: rtk git diff --no-compact]
Also fix --no-compact flag being passed to git (causing usage error)
and remove decorative emoji from compact_diff output.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: propagate exit codes in git diff, status+args, commit, and branch (#632)
4 P1 bugs where git exit codes were swallowed:
- git diff: failure silently printed empty stat output
- git status (with args): failure was filtered instead of propagated
- git commit: failure printed "FAILED" but returned Ok(()) breaking pre-commit hooks
- git branch (list mode): failure was silently ignored
All now follow the established pattern: eprint stderr, track raw==raw, process::exit(code).
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* feat: add 5 new TOML filters (ollama, nx, gradle, spring-boot, jira) (#635)
* feat: add 5 new TOML built-in filters (ollama, nx, gradle, spring-boot, jira)
New filters for commands not covered by Rust modules:
- ollama: strip ANSI spinners, keep final text response (#624)
- nx: strip Nx monorepo noise, keep build results (#444)
- gradle/gradlew: strip UP-TO-DATE tasks, keep build summary (#147)
- spring-boot: strip banner and verbose logs, keep startup/errors (#147)
- jira: strip blanks, truncate wide columns (#524)
All 5 filters pass inline tests via rtk verify (123/123).
Updated builtin filter count: 47 -> 52.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* feat: add 5 more TOML filters (turbo, mise, just, task, yadm)
New filters for task runners and git wrapper:
- turbo: strip cache/Tasks/Duration noise, keep task output (#531)
- mise: strip install/download progress, keep task results (#607)
- just: strip blanks and recipe headers, keep output (#607)
- task: strip task headers and up-to-date lines, keep results (#607)
- yadm: strip hint lines, compact git-like output (#567)
All verified with fake binaries through catch-all TOML engine.
137/137 TOML tests pass, 934 Rust tests pass.
Updated builtin filter count: 52 -> 57.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
---------
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: replace emojis with plain text in git status output (#603) (#638)
Git status output used emojis (📌, 📝, ❓, ✅, ⚠️) that confuse
non-Claude LLMs (GPT, etc.) causing retry loops. Replace with plain
text labels (branch:, modified:, staged:, untracked:, conflicts:).
Also add "clean — nothing to commit" when working tree is clean,
so LLMs understand the repo state without ambiguity.
Before: 📌 master
After: branch: master
clean — nothing to commit
Fixes#603
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
---------
Signed-off-by: Patrick <patrick@rtk.ai>
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
Signed-off-by: Florian Bruniaux <florian@bel-etage.com>
Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>
Signed-off-by: Patrick Szymkowiak <patrick@rtk-ai.app>
Signed-off-by: Ousama Ben Younes <benyounes.ousama@gmail.com>
Signed-off-by: Patrick <patrick@rtk-ai.com>
Co-authored-by: Florian BRUNIAUX <florian@bruniaux.com>
Co-authored-by: Ben Younes <benyounes.ousama@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* security: trust boundary for project-local TOML filters (SA-2025-RTK-002)
.rtk/filters.toml was loaded silently from CWD with highest priority,
allowing an attacker to commit malicious filters to a public repo that
hide security vulnerabilities or rewrite command output via replace and
match_output primitives. CVSS 7.0.
Implementation:
- New src/trust.rs: SHA-256 trust store in ~/.local/share/rtk/
- Trust gate in toml_filter::load(): untrusted filters are SKIPPED
- rtk trust: review + trust project-local filters (stores hash)
- rtk untrust: revoke trust
- rtk trust --list: show all trusted projects
- Content change detection: modified filters auto-rejected
- RTK_TRUST_PROJECT_FILTERS=1 env var override for CI pipelines
- 9 unit tests covering all trust states
Security model: untrusted = skip (not "warn and load"). Zero attack
surface when untrusted. Re-review required on any content change.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* security: harden trust boundary after code review audit
Address 3 critical and 4 important findings from security review:
Critical fixes:
- rtk verify: add trust gate before reading project-local filters
(was executing untrusted filter regex via collect_test_outcomes)
- TOCTOU: read .rtk/filters.toml once in rtk trust, hash in-memory
buffer (prevents file swap between display and hash storage)
- canonical_key: remove fallback path join, fail-closed if
canonicalize fails (prevents symlink injection)
Important fixes:
- RTK_TRUST_PROJECT_FILTERS=1 now requires CI env var co-check
(CI, GITHUB_ACTIONS, GITLAB_CI, JENKINS_URL, BUILDKITE)
to prevent .envrc injection attacks
- Corrupt trust store now logs warning instead of silent fallback
- trusted_at date display uses safe .get(..10) instead of panicking slice
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: rtk untrust graceful when .rtk/filters.toml missing
Previously rtk untrust failed with a canonicalize error if the filter
file had been deleted. Now falls back gracefully with "No trust entry
found" message.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
---------
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* fix: git log --oneline regression drops commits
When user passes --oneline, --pretty, or --format, RTK correctly skips
injecting its custom format with ---END--- markers. However,
filter_log_output() still tried to split on ---END---, treating the
entire output as a single block and only preserving 2 lines.
Add user_format parameter to filter_log_output: when true, use simple
line-based truncation instead of ---END--- block splitting.
Fixes regression introduced in PR #546 (commit c3416eb).
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
* chore: add call-site analysis rule to code-reviewer agent
After the git log --oneline regression (PR #546), add a mandatory
review step: trace all callers of modified functions and verify every
input variant has a test. This prevents regressions where a function
is changed but only one code path is tested.
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
---------
Signed-off-by: Patrick szymkowiak <patrick.szymkowiak@innovtech.eu>
aesoft