Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Skill for the tools module — discovery, validation, scope, and private-sidecar symlink sync for the top-level tools/ directory (executable entry points such as scripts, skills, and agents stored as passive manifests with scripts/ directories). Use when discovering tools (discover_tools), resolving a tool path (resolve_tool_root), validating tool structure (validate_tool_structure), building ToolInfo dataclasses (build_tool_info), checking PUBLIC_TOOL_NAMES for CI-safe scope (public_scope.py), sy
.claude/skills/docxology-infrastructure-tools/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -25% | 0% |
Discovery, validation, scope, and sidecar symlink sync for the tools/ workspace — executable entry points (scripts, skills, agents), distinct from projects/ (executable research code) and fonds/ (passive data stores).
| File | Exports | |------|---------| | __init__.py | ToolInfo, build_tool_info, discover_tools, resolve_tool_root, validate_tool_structure | | discovery.py | discover_tools, resolve_tool_root, NON_RENDERED_TOOL_SUBDIRS, RENDERED_TOOL_SUBDIRS | | tools_info.py | ToolInfo, build_tool_info | | validation.py | validate_tool_structure | | public_scope.py | PUBLIC_TOOL_NAMES, public_tool_infos, public_tool_names | | linking.py | sync_private_tool_links, sync_active_tool_links, private_tools_root, is_managed_symlink, LIFECYCLE_LINK_DIRS, LinkSyncResult |
pythonfrom pathlib import Path from infrastructure.tools import discover_tools, resolve_tool_root, validate_tool_structure, ToolInfo # Discover all valid tools (skips working/, archive/) tools: list[ToolInfo] = discover_tools(Path(".")) for t in tools: print(t.qualified_name, t.tool_type, t.is_valid) # Resolve by name path = resolve_tool_root(Path("."), "templates/template_code_executor") path = resolve_tool_root(Path("."), "template_code_executor") # bare name # Validate structure (requires tools.yaml + scripts/) is_valid, message = validate_tool_structure(Path("tools/templates/template_code_executor"))
pythonfrom infrastructure.tools import ToolInfo, build_tool_info info = build_tool_info(Path("tools/templates/template_code_executor")) info.name # "template_code_executor" info.tool_type # "type" from tools.yaml, or "generic" info.has_scripts # True iff scripts/ exists info.is_valid # True iff has_scripts info.qualified_name # "template_code_executor" or "program/name"
pythonfrom infrastructure.tools.public_scope import ( PUBLIC_TOOL_NAMES, # ('templates/template_code_executor',) public_tool_infos, # list[ToolInfo] present in checkout public_tool_names, # list[str] sorted )
pythonfrom pathlib import Path from infrastructure.tools.linking import ( sync_private_tool_links, private_tools_root, is_managed_symlink, LinkSyncResult, ) root = private_tools_root(Path(".")) # Path | None result: LinkSyncResult = sync_private_tool_links(Path("."), dry_run=True) print(result.summary()) # result.created / .updated / .removed / .skipped / .changed
bashuv run python -m infrastructure.tools.linking --dry-run uv run python -m infrastructure.tools.linking uv run python -m infrastructure.tools.linking --no-prune uv run python -m infrastructure.tools.linking --private-root /path/to/tools
| Path | Required? | |------|-----------| | tools.yaml | ✅ manifest | | scripts/ | ✅ executable entry points | | tests/ | optional | | docs/ | optional |
pythonfrom infrastructure.tools.discovery import NON_RENDERED_TOOL_SUBDIRS, RENDERED_TOOL_SUBDIRS # NON_RENDERED_TOOL_SUBDIRS = frozenset({"working", "archive"}) # RENDERED_TOOL_SUBDIRS = frozenset({"templates"})
| Variable | Purpose | |----------|---------| | TEMPLATE_TOOLS_ROOT | Override private tools root (highest precedence) | | TEMPLATE_SKIP_TOOL_LINK_SYNC | Set to skip auto-sync in orchestration CLI |
PROTECTED_NAMES (public exemplars from PUBLIC_TOOL_NAMES) are always skipped.AGENTS.md — full operating contract, boundaries, lifecycle layout../fonds/SKILL.md — closest analog (fonds module skill)../project/SKILL.md — canonical project-layer skill (same pattern)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→pass | 14,051 | 4,275 | -70% | 1 | 1 | 0% | 3,122 | 2,020 | -35% | 0 | 0 | — |
case-01 | fail→pass | 24,851 | 4,259 | -83% | 1 | 1 | 0% | 2,399 | 2,181 | -9% | 0 | 0 | — |
case-03 | fail→pass | 27,849 | 5,198 | -81% | 1 | 1 | 0% | 3,094 | 2,304 | -26% | 0 | 0 | — |
case-04 | fail→pass | 9,184 | 3,885 | -58% | 1 | 1 | 0% | 1,590 | 2,009 | +26% | 0 | 0 | — |
case-05 | fail→pass | 13,257 | 3,094 | -77% | 1 | 1 | 0% | 2,455 | 1,838 | -25% | 0 | 0 | — |
case-06 | fail→pass | 10,987 | 2,664 | -76% | 1 | 1 | 0% | 1,978 | 1,669 | -16% | 0 | 0 | — |
case-12 | fail→pass | 12,605 | 3,908 | -69% | 1 | 1 | 0% | 2,416 | 2,004 | -17% | 0 | 0 | — |
case-07 | fail→pass | 9,498 | 2,322 | -76% | 1 | 1 | 0% | 1,604 | 1,613 | +1% | 0 | 0 | — |
case-08 | fail→pass | 8,947 | 2,099 | -77% | 1 | 1 | 0% | 1,467 | 1,566 | +7% | 0 | 0 | — |
case-09 | fail→pass | 22,935 | 2,518 | -89% | 1 | 1 | 0% | 2,404 | 1,602 | -33% | 0 | 0 | — |
case-10 | fail→pass | 7,693 | 2,036 | -74% | 1 | 1 | 0% | 1,391 | 1,478 | +6% | 0 | 0 | — |
case-11 | fail→pass | 14,902 | 1,497 | -90% | 1 | 1 | 0% | 2,915 | 1,443 | -50% | 0 | 0 | — |
case-13 | fail→pass | 10,112 | 2,884 | -71% | 1 | 1 | 0% | 1,819 | 1,607 | -12% | 0 | 0 | — |
case-14 | fail→pass | 9,996 | 3,589 | -64% | 1 | 1 | 0% | 1,657 | 1,732 | +5% | 0 | 0 | — |
case-15 | pass→pass | 7,777 | 2,866 | -63% | 1 | 1 | 0% | 1,369 | 1,691 | +24% | 0 | 0 | — |
case-16 | fail→pass | 11,017 | 3,530 | -68% | 1 | 1 | 0% | 1,837 | 1,904 | +4% | 0 | 0 | — |
case-22 | pass→pass | 7,413 | 5,299 | -29% | 1 | 1 | 0% | 1,509 | 2,189 | +45% | 0 | 0 | — |
case-17 | fail→pass | 11,770 | 3,366 | -71% | 1 | 1 | 0% | 2,298 | 1,819 | -21% | 0 | 0 | — |
case-18 | fail→pass | 13,158 | 5,387 | -59% | 1 | 1 | 0% | 2,622 | 2,221 | -15% | 0 | 0 | — |
case-19 | fail→pass | 11,835 | 2,618 | -78% | 1 | 1 | 0% | 2,218 | 1,697 | -23% | 0 | 0 | — |
case-20 | pass→fail | 13,967 | 5,295 | -62% | 1 | 1 | 0% | 3,053 | 1,449 | -53% | 0 | 0 | — |
case-21 | pass→pass | 15,143 | 19,723 | +30% | 1 | 1 | 0% | 2,985 | 5,326 | +78% | 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. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 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 +77 percentage points is the difference between those two pass rates over the 21 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.