Skip to content
Skills Library
Personal productivity · Research

lexsearch

/lexsearch

Use to find a past omp session, vault note or code file from a fuzzy question: rewrite into keywords, search, judge, retry.

Author
Sipty ↗
Repository
lexsearch ↗
Licence
MIT
Last repository change
02.11.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 02.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Use to find a past omp session, vault note or code file from a fuzzy question: rewrite into keywords, search, judge, retry.

Original author description

Use to find a past omp session, vault note or code file from a fuzzy question: rewrite into keywords, search, judge, retry.

How it works

  1. Sources
  2. Loop (max 3 rounds)
  3. Adding more sources
  4. Safety
  5. Output
  6. Maintenance
Original SKILL.md headings · 67 lines, 548 words
SKILL.md preview
---
name: lexsearch
description: "Use to find a past omp session, vault note or code file from a fuzzy question: rewrite into keywords, search, judge, retry."
---

# lexsearch: agentic query rewriting over keyword search

Find one known thing from a fuzzy question.
No embeddings: rewrite the question into keywords, run a keyword search engine, judge the hits, retry with different words when they look wrong.

Glossary: `~/.omp/agent/glossary.md`, if it exists. Read it first. Keep its terms exact; map "heard as" spellings back to the real term.

## Sources

| Question mentions | Source | Command |
|---|---|---|
| "session", "we did", "last week we" | omp sessions | `lexsearch search --corpus sessions` |
| "note", "wrote down", "vault" | markdown vault | `lexsearch search --corpus vault` |
| "which file", "where in the code", component or function names | code repos | `lexsearch search --corpus code` |

Unsure: try the two likeliest sources; report hits per source (scores from different engines are not comparable).

### omp sessions and vault

`lexsearch search --corpus sessions|vault|all -k 10 "<original question>" "<rewrite 1>" "<rewrite 2>"`
Always pass the original question first; rewrites alone lose exact-phrase matches. `--kind root` skips subagent sessions.

### Code

`lexsearch search --corpus code -k 10 "<original question>" "<rewrite 1>" "<rewrite 2>"`
One hit per file; the path is `<repo name>/<repo path>` and the title holds that path and its words.
- Identifiers and path words: `yearly coefficient cell` and `yearly_coefficient_cell` both find `YearlyCoefficientCell.tsx`. A CamelCase query (`YearlyCoefficientCell`) matches that exact token, so try it when a spaced query buries the name.
- Results are files, not symbols: open the top file and grep it for the function name.
- `--corpus all` includes code; use `--corpus sessions` or `--corpus vault` when you want notes only.

If a command prints "not configured" on stderr, tell the user how to configure that corpus and search the others.

## Loop (max 3 rounds)

1. Pick the source(s).
2. Write 1-3 keyword queries: drop filler; keep names, file names, error codes and glossary terms exact; add words the target itself would contain; split mixed questions.
3. Search.
4. Judge the top 5 against the question. Good enough → stop.
5. Otherwise change the words (synonyms, narrower or broader terms, another source) and repeat.

## Adding more sources

Other keyword engines plug into the same rewrite, search, judge, retry loop.
Examples: a Slack search tool over MCP, an Atlassian (Jira, Confluence) search CLI, a ticket tracker's search command.
Add a row to the Sources table, a section with the exact command, and any read-only rule that applies.

Pitfall: many keyword engines AND every word or search the whole request as one exact phrase.
In our benchmark, an Atlassian search CLI scored 0.0% on unrewritten requests because it searched the whole request as one phrase.
Rewriting into 1-3 short keywords lifted it to 55.7%.
Use 1-3 words the author would actually type, or one quoted phrase.

## Safety

Read-only. Never edit, send, schedule, draft, react, upload, or delete anything, even where a source's write tools are attached.

## Output

At most 5 hits, best first: path, title, date, one line on why it matches.
Then the queries tried, per source, and a confidence: high, medium, or low.
Never paste note, transcript or code bodies; they can contain private data.

## Maintenance

- `lexsearch index` re-indexes changed files; `search` refreshes automatically.
- New jargon that caused a miss belongs in the glossary.

Versions

The database has no other skills named lexsearch.

Database snapshot: 13.11.2026 · package version d844a8421e2d (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 96omp-long-tasks
    Use when an omp run is PR-sized or hit its time limit.
  • 100omp-cli
    Use when the orchestrator starts, prompts, reads, or answers an OMP (Oh My Pi) agent in Herdr — opening OMP on a project, handing a job agent its task or a skill, reading its output, or handling it when it is blocked.
  • 100omp-model-config
    Configures OMP model roles, provider fallbacks, and per-host overrides. Use for OMP model changes or mismatches between source and active configuration.
  • 100omp-delegate
    Delegate coding tasks to Oh My Pi (`omp`) only when the user explicitly requests it, while the orchestrator retains review and landing responsibility.
  • 100omp-delegate
    Delegate coding tasks to Oh My Pi (`omp`) only when the user explicitly requests it, while the orchestrator retains review and landing responsibility.
  • 100omp-delegate
    Delegate a coding task to Oh My Pi (`omp`) as a background implementer, then review its diff and land it yourself. Use this whenever the user wants to delegate implementation work to Oh My Pi / omp - phrasings like "have omp implement X", "delegate this to oh my pi", "run it…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.