ee1df22 narrowed the Claude Code search-guard matcher from "Glob|Grep" to
"Bash" on the premise that dedicated search tools were removed and searches
go through Bash. Current Claude Code routes content search through its
first-class Grep tool (its Bash tool description actively steers away from
shell grep), so the graphify-first nudge never fired on the agent's primary
exploration path and the graph was silently bypassed.
Three-part fix, per the issue's analysis:
- Matcher: "Bash" -> "Bash|Grep" in _claude_pretooluse_hooks. Glob already
fires the read nudge via "Read|Glob", so Grep was the only orphaned tool.
- Guard body: the hook-guard search branch only inspected tool_input.command,
which a Grep call doesn't carry (it has pattern/path/glob). A Grep-shaped
input (pattern present, no command) is now treated as a search — it IS one
by definition — and nudges whenever a fresh graph exists. The Bash
token-matching path is unchanged, and a command-carrying input never
triggers the Grep shape, so non-search Bash calls stay silent.
- Idempotency: "Bash|Grep" added to the four install/uninstall dedup filters
(claude + codebuddy), so upgrading replaces the stale "Bash" hook in place
instead of appending a duplicate — verified against a pre-fix settings.json.
Tests: new regression tests feed Grep-shaped tool_input through
hook-guard search and assert the nudge (with graph), silence (without),
valid PreToolUse JSON, and no blocking; plus a guard that a non-search Bash
command with a stray pattern key does not nudge. Existing matcher assertions
updated across test_search_hook/test_install/test_claude_md/test_codebuddy/
test_hook_strict. Hook+install suites: 397 passed. Full suite: 3224 passed;
the 13 failures are pre-existing on clean v8 in this environment.
Fixes#1986
Claude Code runs command-type PreToolUse hooks through Git Bash by default on
Windows. The resolved exe was emitted as a raw backslash path and quoted only
when it contained a space, so a space-free path like C:\Users\me\graphify.EXE
reached settings.json unquoted. Git Bash treats the unquoted backslashes as
escapes and strips them, producing 'C:Usersmegraphify.EXE: command not found',
so every graph guard silently fails. Normalize the path to forward slashes in
_resolve_graphify_exe (a no-op on POSIX), fixing the Claude, Gemini, and Codex
hook emitters at once.
Agents routinely ignore the advisory "run graphify query first" nudge and read
raw files anyway. `graphify install --project --strict` (or `graphify claude
install --strict`) now installs a hook that BLOCKS the first raw source read of a
session via permissionDecision:"deny" with a redirect to graphify query, then
downgrades to the soft nudge — it fires at most once per session (atomic
per-session marker) so it can never strand the agent, and a recent
query/explain/path refreshes a stamp that suppresses it. Claude Code only;
Bash-grep and Glob stay nudge-only; Gemini/Codex/OpenCode are unchanged.
GRAPHIFY_HOOK_STRICT=1/0 toggles at runtime without a reinstall; default installs
are byte-identical (soft nudge).
Also fixes#1840 for the default soft nudge: the guard no longer fires for reads
of out-of-project files, and softens to a non-mandatory nudge when the graph is
stale for the target file. Gating is ~3 stat calls (no corpus walk) and fails
open. Begins the 0.9.19 cycle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
main() carried a 256-line if/elif chain routing every install-family command
(install, uninstall, and the 22 per-platform targets: claude, codebuddy, gemini,
cursor, vscode, copilot, kilo, kiro, devin, pi, amp, agents/skills,
aider/codex/opencode/claw/droid/trae/trae-cn/hermes, antigravity). That block
only ever reads sys.argv and calls functions that already live in install.py, so
move it there verbatim as a guarded dispatch_install_cli(cmd) -> bool; main() now
does `if dispatch_install_cli(cmd): return`.
The command set is derived from the block's own conditions (so none is missed),
and the branch's early `return` becomes `return True` so the help path still
short-circuits. Verified end-to-end: `graphify claude install/uninstall`, unknown
command, and the install test suite all behave identically.
__main__.py 3,641 -> 3,388 LOC (5,368 at branch start, -37%). Full suite
unchanged: 3036 passed, 29 skipped; skillgen --check OK.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
__main__.py was 5,368 LOC, more than half of it per-platform install/uninstall
machinery interleaved with the CLI dispatcher. Move that subsystem — 68
functions (all *_install/*_uninstall, _copy_skill_file, _platform_skill_destination,
_always_on, etc.) plus 21 module constants (_PLATFORM_CONFIG and the platform
skill/hook payload constants) — into a new graphify/install.py, extracted verbatim.
Behavior-preserving:
- install.py lives in the same package dir as __main__ so packaged-asset lookups
via Path(__file__).parent ("always_on"/"skills") resolve unchanged.
- __main__ re-exports all 89 moved names, so `from graphify.__main__ import
claude_install` (and every other import, incl. private helpers used by tests)
keeps working, with object identity preserved.
- The install cluster was verified (AST) to have zero back-references into
__main__, so no circular import; __main__ imports _PLATFORM_CONFIG et al. one-way.
__main__.py drops 5,368 -> 3,642 LOC. Full suite unchanged: 3036 passed, 29
skipped (excluding the env-only openai tests). ruff check clean; skillgen --check OK.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
shazeb