tim-hub/dig-knowledge

Dig Knowledge

Search the knowledge base for past development context — what was done, why, and what challenges were faced — to inform current work.

This skill is the read side of the knowledge system. It pulls in relevant history so you don't repeat mistakes or re-discover solutions that were already found.

Silent Skip Rule

If any script in Step 1 exits non-zero, skip silently — do not warn, prompt, or mention the knowledge base. This skill should never be a friction point. If there's nothing to read, move on.

Step 1: Load Config and Resolve Project

Run the load script:

python3 "${CLAUDE_PLUGIN_ROOT}/skills/dig-knowledge/scripts/kb-load.py" "[optional_project_name]"

• Pass the user's argument as the project name if it looks like a project name (single word, no spaces).
• If no argument or the argument is a search query, omit it — the script derives the project from the cwd.
• If the script exits non-zero → silent skip. Config missing, path not set, or folder doesn't exist.
• If it exits 0, it prints JSON: { "kb_path": "...", "project": "...", "project_path": "...", "has_entries": true/false }
• If has_entries is false → silent skip.

Step 2: Search Knowledge

Run the search script:

python3 "${CLAUDE_PLUGIN_ROOT}/skills/dig-knowledge/scripts/kb-search.py" "{project_path}" "[search_term]"

• If the user provided a search term (e.g., "auth issues", "CSV export bug"), pass it as the second argument. The script matches filenames first, then greps content.
• If no search term, omit it — the script returns the 5 most recent entries.
• The script outputs one file path per line. Read the returned files.

Step 3: Present Findings

When invoked explicitly by the user — show a structured summary:

• List matching entries with dates and titles
• Show the key bullet points from each relevant entry
• Highlight anything that might be relevant to the current task

When invoked silently during explore/plan — inject context naturally:

• Don't announce "I found knowledge base entries"
• Instead, weave relevant history into your exploration or planning output
• Reference past decisions where they inform current choices (e.g., "Previously, X approach was chosen because Y — this is still relevant here")

Integration with Explore and Plan

When the explore or plan skills run, this skill can be invoked as a preliminary step to load project history. The calling skill should:

1. Call dig-knowledge with the current project name
2. If results come back, include them as context for the exploration/plan
3. If silent skip, proceed normally

This integration is optional — explore and plan work fine without it. Knowledge just makes them better informed.