Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Find source code with Memtrace hybrid BM25+semantic search: symbols, functions, classes, types, constants, definitions, implementations, logic, or error strings inside code. Use when the user wants to find, search, locate, or look up code or asks where code lives. Do not use Grep, Glob, rg, find, or manual file search for code discovery. If Memtrace returns 0 results, broaden the Memtrace query and diagnose/reindex; do not switch to grep.
.claude/skills/syncable-dev-memtrace-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 262% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 109% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 95% | 0% |
| case-21 | ✓→✗ | ▼ Worse | 96% | 0% |
Find code using hybrid BM25 + semantic search (RRF). Primary discovery tool — use before relationship or impact analysis.
| Tool | Best For | |------|----------| | find_code | Natural-language queries, broad searches | | find_symbol | Exact identifier names | | get_source_window | Bounded source read when harness lacks Read(offset, limit) |
> Parameter types: Numbers must be JSON numbers — not strings.
find_code parameters| Param | Required | Default | Notes | |---|---|---|---| | query | yes | — | Natural language or symbol text | | repo_id | no | all repos | | | limit | no | 20 | Max 100 | | file_path | no | — | Path/directory substring filter | | as_of | no | now | ISO-8601 time-travel | | include_diagnostics | no | false | Set true for id, score in results |
No kind param on find_code — use find_symbol(kind=...) to filter by symbol type.
json{ "query": "authentication middleware", "repo_id": "memdb", "limit": 20 }
find_symbol parameters| Param | Required | Default | Notes | |---|---|---|---| | name | yes | — | Symbol name to search | | repo_id | no | all repos | | | fuzzy | no | false | API field exists; currently exact-match in backend | | edit_distance | no | 2 | Only when fuzzy enabled | | kind | no | — | Function, Class, Method, etc. | | file_path | no | — | Path substring filter | | limit | no | 10 | Max 50 |
json{ "name": "validateToken", "repo_id": "memdb", "file_path": "auth" }
Full parameter spec for every Memtrace tool: references/mcp-parameters.md (bundled at the memtrace-skills plugin root).
find_symbolfind_coderepo_idResult shape: see Output below.
Save name, scope_path, and file_path — not internal IDs:
json{ "repo_id": "memdb", "symbol": "validateToken" } { "repo_id": "memdb", "target": "validateToken", "direction": "both" } { "repo_id": "memdb", "target": "validateToken", "query_type": "find_callers" }
Read source only when editing — bounded Read(offset, limit) at returned lines.
find_code query.file_path:start_line.One find_code / find_symbol result entry:
json{ "name": "validateToken", "kind": "Function", "file_path": "src/auth/token.rs", "start_line": 42, "scope_path": "auth::token" }
score (and id) appear only with include_diagnostics: true.
| Mistake | Reality | |---------|---------| | find_code(kind=...) | kind only on find_symbol | | Passing symbol id to graph tools | Use name as symbol / target | | Assuming fuzzy: true always works | Backend is exact-match today — try spelling variants | | Skipping list_indexed_repositories | Verify repo is indexed first |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-19 | fail→fail | 13,085 | 5,986 | -54% | 1 | 1 | 0% | 2,037 | 1,937 | -5% | 0 | 0 | — |
case-03 | fail→fail | 26,053 | 6,174 | -76% | 1 | 1 | 0% | 181 | 1,205 | +566% | 0 | 0 | — |
case-14 | fail→fail | 10,483 | 3,815 | -64% | 1 | 1 | 0% | 1,600 | 1,146 | -28% | 0 | 0 | — |
case-01 | fail→fail | 8,166 | 4,521 | -45% | 1 | 1 | 0% | 1,368 | 1,194 | -13% | 0 | 0 | — |
case-02 | fail→fail | 12,843 | 6,639 | -48% | 1 | 1 | 0% | 1,469 | 1,240 | -16% | 0 | 0 | — |
case-04 | fail→pass | 11,466 | 15,742 | +37% | 1 | 1 | 0% | 611 | 2,214 | +262% | 0 | 0 | — |
case-05 | pass→pass | 4,111 | 3,661 | -11% | 1 | 1 | 0% | 708 | 1,586 | +124% | 0 | 0 | — |
case-06 | fail→fail | 10,406 | 3,736 | -64% | 1 | 1 | 0% | 1,717 | 1,169 | -32% | 0 | 0 | — |
case-07 | fail→pass | 4,442 | 3,737 | -16% | 1 | 1 | 0% | 817 | 1,707 | +109% | 0 | 0 | — |
case-08 | fail→pass | 10,178 | 6,091 | -40% | 1 | 1 | 0% | 1,712 | 2,055 | +20% | 0 | 0 | — |
case-09 | fail→pass | 4,000 | 2,941 | -26% | 1 | 1 | 0% | 726 | 1,413 | +95% | 0 | 0 | — |
case-10 | fail→fail | 8,310 | 5,165 | -38% | 1 | 1 | 0% | 444 | 1,219 | +175% | 0 | 0 | — |
case-11 | fail→fail | 11,006 | 3,990 | -64% | 1 | 1 | 0% | 1,971 | 1,155 | -41% | 0 | 0 | — |
case-12 | fail→fail | 8,229 | 4,487 | -45% | 1 | 1 | 0% | 1,283 | 1,146 | -11% | 0 | 0 | — |
case-13 | fail→fail | 15,849 | 4,725 | -70% | 1 | 1 | 0% | 2,327 | 1,189 | -49% | 0 | 0 | — |
case-15 | pass→pass | 6,418 | 2,294 | -64% | 1 | 1 | 0% | 1,026 | 1,315 | +28% | 0 | 0 | — |
case-16 | pass→pass | 17,276 | 3,009 | -83% | 1 | 1 | 0% | 2,815 | 1,436 | -49% | 0 | 0 | — |
case-17 | fail→fail | 6,838 | 2,173 | -68% | 1 | 1 | 0% | 1,045 | 1,248 | +19% | 0 | 0 | — |
case-18 | pass→pass | 5,321 | 3,242 | -39% | 1 | 1 | 0% | 890 | 1,463 | +64% | 0 | 0 | — |
case-20 | fail→fail | 14,022 | 10,583 | -25% | 1 | 1 | 0% | 2,305 | 1,737 | -25% | 0 | 0 | — |
case-21 | pass→fail | 6,219 | 14,106 | +127% | 1 | 1 | 0% | 1,098 | 2,149 | +96% | 0 | 0 | — |
case-22 | fail→fail | 19,264 | 7,429 | -61% | 1 | 1 | 0% | 3,156 | 1,312 | -58% | 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 10 counted toward the lift figure. The other 12 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 +14 percentage points is the difference between those two pass rates over the 10 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.