Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use terminal_rg / terminal_find when you need raw filesystem search outside the project tree — system configs, /var/log, /etc, archive contents — or when files-tools.search_files is too project-scoped. Teaches the rg vs find vs terminal_exec("ls/du/tree") split, common rg flag combos for code/logs/configs, find predicates for mtime/size/type queries, and the rule that for tree views or single-file stat info you should just use terminal_exec instead of inventing a tool. Read before reaching for r
.claude/skills/aden-hive-hive-terminal-tools-fs-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 182% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 48% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 292% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 102% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 49% | 0% |
terminal-tools provides two structured search tools: terminal_rg (ripgrep for content) and terminal_glob (find files by name/glob). Predicate queries (mtime/size/type) and everything else (tree, stat, du) are just terminal_exec.
| Task | Tool | |---|---| | Find code/text matching a pattern (project tree or any path) | terminal_rg (gitignore-aware; defaults to your session workdir) | | Find files by name/glob (any path) | terminal_glob | | Find files by mtime/size/type predicate | terminal_exec("find ...") (see references/find_predicates.md) | | List a directory | terminal_exec("ls -la /path") | | Tree view | terminal_exec("tree -L 2 /path") | | Single-path stat | terminal_exec("stat /path") | | Disk usage | terminal_exec("du -sh /path") or terminal_exec("du -h --max-depth=2 /") | | Count matches across files | terminal_rg(pattern, count=True via extra_args=["-c"]) |
terminal_rg — content searchripgrep is fast, gitignore-aware, and has a deep flag surface. The structured wrapper exposes the most useful flags directly; extra_args covers the rest.
# All Python files containing "TODO"
terminal_rg(pattern="TODO", path=".", type_filter="py")
# Case-insensitive, with context
terminal_rg(pattern="error", path="/var/log", ignore_case=True, context=2)
# Search hidden files (rg ignores them by default)
terminal_rg(pattern="api_key", path="~", hidden=True)
# Don't respect .gitignore (find files git would ignore)
terminal_rg(pattern="generated", path=".", no_ignore=True)
# Multi-line pattern (e.g., function definitions spanning lines)
terminal_rg(pattern=r"def\s+\w+\(.*\n.*\n", path="src", extra_args=["--multiline"])
# Specific filename glob
terminal_rg(pattern="version", path=".", glob="*.toml")| Flag | Effect | |---|---| | -tpy (type_filter="py") | Only Python files | | -uu | Don't respect any ignores (incl. .git/) | | --multiline (extra_args) | Allow regex spanning lines | | --max-count (max_count) | Stop after N matches per file | | --max-depth (max_depth) | Limit recursion | | -w (extra_args) | Whole word match | | -F (extra_args) | Fixed string (no regex) |
See references/ripgrep_cheatsheet.md for the long form.
terminal_glob — find files by nameLists files matching a glob, gitignore-aware (backed by rg --files). The pattern is widened for you so a bare stem Just Works — the actual glob run is returned as expanded_pattern:
lk_scan_post_reactors → matched as **/*lk_scan_post_reactors* (recursive substring)*.py → matched as **/*.py (recursive by default)src/**/*.py → used verbatim# Find a file by stem anywhere under a tree
terminal_glob(pattern="lk_scan_post_reactors", path="core/framework/skills")
# All YAML configs under /etc
terminal_glob(pattern="*.yaml", path="/etc")
# Include .gitignored / hidden / build-cache files
terminal_glob(pattern="*.log", path=".", include_ignored=True)For predicate queries (modified in last N days, larger than N MB, only dirs/symlinks), terminal_glob is the wrong tool — drop to terminal_exec("find ..."). See references/find_predicates.md.
Both tools return truncated: true when output exceeded the inline cap. For terminal_rg, matches were dropped (refine the pattern or narrow the path); for terminal_glob, results past max_results (default 1000) were dropped — the search stops early at the cap, so narrow the pattern rather than raising it. terminal_glob also returns timed_out: true (with the partial results it gathered) when the walk exceeded its deadline.
terminal_rg is the project search tool — gitignore-aware and returns structured matches; use it for in-project search as well as raw paths.terminal_glob to list one directory — terminal_exec("ls -la /path") is shorter.terminal_exec("grep ...") when terminal_rg exists — rg is faster, gitignore-aware, and returns structured matches.terminal_exec("find ... -name ...") for a plain name search — use terminal_glob. Reserve terminal_exec("find ...") for mtime/size/type predicates.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,324 | 6,065 | -17% | 1 | 1 | 0% | 1,384 | 1,608 | +16% | 0 | 0 | — |
case-02 | fail→fail | 16,346 | 7,950 | -51% | 1 | 1 | 0% | 3,171 | 1,779 | -44% | 0 | 0 | — |
case-03 | fail→pass | 4,219 | 4,387 | +4% | 1 | 1 | 0% | 717 | 2,021 | +182% | 0 | 0 | — |
case-04 | fail→fail | 3,874 | 5,066 | +31% | 1 | 1 | 0% | 664 | 1,627 | +145% | 0 | 0 | — |
case-05 | fail→pass | 6,411 | 3,385 | -47% | 1 | 1 | 0% | 1,253 | 1,850 | +48% | 0 | 0 | — |
case-06 | fail→pass | 3,129 | 3,604 | +15% | 1 | 1 | 0% | 474 | 1,857 | +292% | 0 | 0 | — |
case-07 | fail→fail | 4,279 | 4,446 | +4% | 1 | 1 | 0% | 783 | 2,041 | +161% | 0 | 0 | — |
case-08 | fail→pass | 4,977 | 4,111 | -17% | 1 | 1 | 0% | 962 | 1,948 | +102% | 0 | 0 | — |
case-09 | fail→pass | 5,800 | 2,140 | -63% | 1 | 1 | 0% | 1,096 | 1,632 | +49% | 0 | 0 | — |
case-10 | fail→pass | 8,899 | 5,718 | -36% | 1 | 1 | 0% | 1,676 | 2,350 | +40% | 0 | 0 | — |
case-11 | fail→pass | 3,878 | 2,180 | -44% | 1 | 1 | 0% | 687 | 1,647 | +140% | 0 | 0 | — |
case-12 | fail→pass | 4,707 | 3,708 | -21% | 1 | 1 | 0% | 885 | 1,937 | +119% | 0 | 0 | — |
case-13 | fail→fail | 5,469 | 5,108 | -7% | 1 | 1 | 0% | 1,168 | 1,528 | +31% | 0 | 0 | — |
case-14 | fail→fail | 4,702 | 5,439 | +16% | 1 | 1 | 0% | 917 | 1,618 | +76% | 0 | 0 | — |
case-15 | fail→pass | 2,828 | 5,969 | +111% | 1 | 1 | 0% | 504 | 1,876 | +272% | 0 | 0 | — |
case-16 | fail→pass | 6,056 | 2,084 | -66% | 1 | 1 | 0% | 1,081 | 1,648 | +52% | 0 | 0 | — |
case-17 | fail→pass | 6,340 | 2,269 | -64% | 1 | 1 | 0% | 1,100 | 1,620 | +47% | 0 | 0 | — |
case-18 | fail→pass | 6,739 | 3,296 | -51% | 1 | 1 | 0% | 1,200 | 1,895 | +58% | 0 | 0 | — |
case-19 | pass→pass | 6,023 | 3,890 | -35% | 1 | 1 | 0% | 991 | 1,887 | +90% | 0 | 0 | — |
case-20 | fail→pass | 4,296 | 3,123 | -27% | 1 | 1 | 0% | 825 | 1,751 | +112% | 0 | 0 | — |
case-21 | pass→pass | 5,821 | 3,266 | -44% | 1 | 1 | 0% | 1,174 | 1,826 | +56% | 0 | 0 | — |
case-22 | pass→fail | 8,806 | 4,942 | -44% | 1 | 1 | 0% | 1,837 | 1,564 | -15% | 0 | 0 | — |
case-23 | pass→pass | 6,324 | 4,458 | -30% | 1 | 1 | 0% | 1,108 | 2,056 | +86% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 17 counted toward the lift figure. The other 6 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +52 percentage points is the difference between those two pass rates over the 17 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.