Claude Skill

ingest

Turn a raw session notes file into an ingested note, ready for the enrich stage. Defaults to keeping it whole; only splits when bullets show a real cross-topic signal. Stage 1 of the processing pipeline.

LLM Mart · 0 points · 1 views 0 listing impressions 0 install-command copies
Download technicalpickles-pickled-claude-plugins-plugins_second-brain_skills_ingest-85f769c.zip · 1 KB

Install

skills CLI npx skills add https://github.com/technicalpickles/pickled-claude-plugins/tree/main/plugins/second-brain/skills/ingest
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install technicalpickles-pickled-claude-plugins@llmmart
Git git clone https://github.com/technicalpickles/pickled-claude-plugins.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole technicalpickles/pickled-claude-plugins collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Ingest Stage

Turn a raw session notes file (type: session-notes) into an ingested insight, ready for the enrich stage. Defaults to keeping it as one note — see "Exception" below for when to split.

See ../obsidian/references/pipeline.md for stage definitions and status flow.

Input

A session notes file in the inbox with status: raw and type: session-notes.

If the note names a Claude Code session (source: claude-code session <id> ... or a session/session-id key) and that session's transcript isn't on this machine, leave the note untouched and say so: it was written on another machine and waits for that one. process-inbox finds these in its survey; when you're run on a single note, check with ls ~/.claude/projects/*/<id>*.jsonl (each CLAUDE_CONFIG_DIR entry in place of ~/.claude when it is set).

Default: enrich whole

Most session notes — including every devlog-produced one — are already well-organized: one session, one provenance (repo/branch/issue), a handful of related bullets. Splitting them into N separate notes mints near-duplicate notes that mostly route to the same destination anyway — a real run against 32 session notes would have exploded into 120-140 near-duplicates that mostly filed to the same destination.

Default behavior: keep the session note as ONE note. Change its frontmatter in place, using the Edit tool:

  • status: raw → status: ingested
  • type: session-notes → type: insight

Don't create any new files. Don't touch the body — enrich handles prose cleanup next.

Exception: split when bullets genuinely diverge

Check for a real cross-topic signal before splitting — not a bullet count. Read the note's bullets:

  • If every bullet shares the same repo/branch/issue context (the common case, and always true for a single devlog session), keep it whole.

  • If bullets explicitly reference different repos, branches, or issues within the same file (only possible in hand-authored or legacy multi-topic session notes — devlog output never does this), split those bullets into separate notes instead:

    There's no structured per-bullet frontmatter for this — repo/branch/issue only ever exist once, at the file level. Infer divergence from what the bullet text itself says (e.g. a bullet's prose names a different repo than the file's own frontmatter, or than another bullet). Older notes may carry bean: instead of issue:; treat them the same, and copy whichever key the source note used.

    1. For each insight, create a new note via sb note create --source auto --title "{insight title}" with status: ingested, type: insight, source-session: {original session filename}, and that bullet's own repo/branch/issue.
    2. Write the note body (the bullet text, cleaned into a sentence or short paragraph) using the Write tool.
    3. Update the original session file's status to ingested using the Edit tool — it stays in the inbox as a record, not processed further.
    4. Skip empty or trivial bullets when splitting (single words, fragments).

Output

Either:

  • The same note, now status: ingested and type: insight (default case), or
  • N new insight notes with status: ingested, plus the original session file updated to status: ingested (split case)

Constraints

  • Default to whole. Splitting is the exception, not the rule.
  • Preserve all provenance metadata.
  • Each split-out note gets its own zettelkasten timestamp (via sb).
Files (pickled-claude-plugins)
  • SKILL.md 3.8 KB
    ---
    name: ingest
    description: Turn a raw session notes file into an ingested note, ready for the enrich stage. Defaults to keeping it whole; only splits when bullets show a real cross-topic signal. Stage 1 of the processing pipeline.
    allowed-tools:
      - Read(~/.claude/vaults/**/*.md)
      - Write(~/.claude/vaults/**/*.md)
      - Edit(~/.claude/vaults/**/*.md)
      - Bash(npx @techpickles/sb:*)
    ---
    
    # Ingest Stage
    
    Turn a raw session notes file (type: `session-notes`) into an ingested
    insight, ready for the `enrich` stage. Defaults to keeping it as one
    note — see "Exception" below for when to split.
    
    See `../obsidian/references/pipeline.md` for stage definitions and status flow.
    
    ## Input
    
    A session notes file in the inbox with `status: raw` and `type: session-notes`.
    
    If the note names a Claude Code session (`source: claude-code session <id> ...`
    or a `session`/`session-id` key) and that session's transcript isn't on this
    machine, leave the note untouched and say so: it was written on another
    machine and waits for that one. `process-inbox` finds these in its survey;
    when you're run on a single note, check with
    `ls ~/.claude/projects/*/<id>*.jsonl` (each `CLAUDE_CONFIG_DIR` entry in
    place of `~/.claude` when it is set).
    
    ## Default: enrich whole
    
    Most session notes — including every `devlog`-produced one — are already
    well-organized: one session, one provenance (repo/branch/issue), a handful
    of related bullets. Splitting them into N separate notes mints
    near-duplicate notes that mostly route to the same destination anyway — a
    real run against 32 session notes would have exploded into 120-140
    near-duplicates that mostly filed to the same destination.
    
    Default behavior: keep the session note as ONE note. Change its
    frontmatter in place, using the Edit tool:
    - `status: raw` → `status: ingested`
    - `type: session-notes` → `type: insight`
    
    Don't create any new files. Don't touch the body — `enrich` handles prose
    cleanup next.
    
    ## Exception: split when bullets genuinely diverge
    
    Check for a real cross-topic signal before splitting — not a bullet count.
    Read the note's bullets:
    
    - If every bullet shares the same repo/branch/issue context (the common
      case, and always true for a single `devlog` session), keep it whole.
    - If bullets explicitly reference **different** repos, branches, or issues
      within the same file (only possible in hand-authored or legacy
      multi-topic session notes — `devlog` output never does this), split
      those bullets into separate notes instead:
    
      There's no structured per-bullet frontmatter for this — repo/branch/issue
      only ever exist once, at the file level. Infer divergence from what the
      bullet text itself says (e.g. a bullet's prose names a different repo
      than the file's own frontmatter, or than another bullet). Older notes
      may carry `bean:` instead of `issue:`; treat them the same, and copy
      whichever key the source note used.
    
      1. For each insight, create a new note via
         `sb note create --source auto --title "{insight title}"` with
         `status: ingested`, `type: insight`, `source-session: {original session filename}`,
         and that bullet's own repo/branch/issue.
      2. Write the note body (the bullet text, cleaned into a sentence or
         short paragraph) using the Write tool.
      3. Update the original session file's status to `ingested` using the
         Edit tool — it stays in the inbox as a record, not processed
         further.
      4. Skip empty or trivial bullets when splitting (single words,
         fragments).
    
    ## Output
    
    Either:
    - The same note, now `status: ingested` and `type: insight` (default
      case), or
    - N new insight notes with `status: ingested`, plus the original session
      file updated to `status: ingested` (split case)
    
    ## Constraints
    
    - Default to whole. Splitting is the exception, not the rule.
    - Preserve all provenance metadata.
    - Each split-out note gets its own zettelkasten timestamp (via sb).
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.