Skip to content
SkillModel-led874 tokens in use

session-search

Answers which past agent session handled a topic by ranking the session-brain graph, then loads the best transcript into your current conversation.

3,792 stars on the repository MIT LicenseAr9av/obsidian-wiki · skills/session-search
pip install obsidian-wiki
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing session-search costs

58tokens idle
874tokens in use

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

shelldetected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter8/15
  • Structure17/25
  • Examples9/15
  • Output5/15
  • Size15/15
  • Resources10/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent
  • Runs shell commands−5

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 14

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Excellent
  • Development activity100
  • Maintenance100
  • Community60
  • Maturity28
  • Momentum40

session-search is a read-only lookup over the graph that session-brain builds. You give it a topic, it queries the graph through the obsidian-wiki command line tool and presents the top matches in a compact table with the reason each one matched.

It then hands the best session ids to a loader, or reads the human turns of the transcript directly, and answers from what it loaded while citing the session ids it used.

What You Get

  • A short table of matching past sessions with title, project, date, topic and match reason
  • Filters for project, cluster, start date and result count
  • The best transcript loaded into the current conversation, read only through its human turns
  • A warning when the graph is older than about a week, with an offer to rebuild
  • An answer that cites the session ids it drew on and says when only prompts survive

Common Use Cases

  • Finding the session where a bug was fixed months ago
  • Checking whether a task was already done before starting it again
  • Pulling the context of a bookmarked session back into a new chat

What Makes It Unique

Ranking uses the cluster a session belongs to, so a session that never used your exact word can still surface for the topic. It separates sessions whose transcripts are gone from loadable ones and tells you how much context is really available. It hands loading to an installed loader skill when there is one and has a built-in fallback that filters a transcript to the human turns, so a huge file is never read whole.

Under The Hood

The skill is instructions that run the obsidian-wiki command line tool through the shell for sessions-query and sessions-show. As a fallback it runs grep to size a transcript and an inline Python snippet that reads the JSONL file and prints the first part of each user message. It reads session files and the graph on disk. It does not write files, fetch web pages or call remote APIs, and it never rebuilds the graph without asking.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers