Context Mine

See also: context-ingest is the single public entry point for new material. If the user dropped a conversation export of unknown shape, prefer context-ingest — it dispatches here automatically when it detects a Slack/transcript/chat-log shape. Use context-mine directly only when the user explicitly names mining ("mine this transcript", "extract decisions from this Slack export"). Mine extracts → ingest classifies and routes.

Purpose

This skill IS:

• An extraction tool that reads conversation exports and pulls out actionable context
• A bridge between unstructured conversations and the CLEAR context architecture
• A preprocessor that dumps structured summaries to docs/_inbox/ for review

This skill IS NOT:

• A replacement for context-ingest (mine extracts, ingest classifies and routes)
• An automatic integration tool (everything goes to _inbox for human review)
• A real-time conversation monitor (works on exports/transcripts after the fact)

When to Use

• "Extract context from this Slack export"
• "Mine this meeting transcript for decisions"
• "Process this chat log into actionable items"
• "I have a conversation dump, what did we decide?"

Supported Input Formats

Extraction Categories

For each conversation, extract into these categories:

Decisions Made

Things that were decided, with enough context to understand why.

- Decision: [what was decided]
  Rationale: [why, if stated]
  Affects: [what data points this might update]

Discoveries

New information learned that wasn't known before.

- Discovery: [what was learned]
  Source: [who said it / where it came from]
  Relevance: [which data point or cluster this relates to]

Action Items

Tasks, follow-ups, commitments.

- [ ] [action item]
  Owner: [who, if mentioned]
  Deadline: [when, if mentioned]

Preferences & Positions

Opinions, stances, preferences expressed by key people.

- Preference: [what was expressed]
  Who: [who said it]
  Context: [in response to what]

Problems & Risks

Issues raised, concerns flagged, blockers identified.

- Problem: [what was flagged]
  Severity: [high/medium/low based on language used]
  Status: [resolved in conversation or still open]

Workflow

Step 1: Receive Input

Accept the conversation in any supported format. If pasted directly, work with it as-is. If a file path, read the file.

Step 2: Scan and Extract

Read through the conversation and extract items into the 5 categories above. Guidelines:

• Be selective — only extract items with clear business relevance
• Preserve attribution — note who said what when possible
• Link to data points — if an extracted item relates to a known data point, note which one
• Skip small talk — ignore greetings, scheduling logistics, off-topic chatter

Step 3: Write to Inbox

Create one or more files in docs/_inbox/ with frontmatter:

---
type: session-capture
source: "[slack-export | meeting-transcript | chat-log | paste]"
date: "YYYY-MM-DD"
status: raw
mined-from: "[filename or 'direct paste']"
---

Step 4: Summarize

Report to the user:

• How many items extracted per category
• Which existing data points are likely affected
• Recommend: "Run context-ingest to classify and route these items"

What AI Does vs What's Mechanical

Example Output

---
type: session-capture
source: slack-export
date: 2026-04-07
status: raw
mined-from: "product-channel-export-april.json"
---

## Decisions
- Switched to usage-based pricing for enterprise tier
  Rationale: Customer interviews showed resistance to per-seat model
  Affects: pricing-strategy.md, value-proposition.md

## Discoveries
- Competitor X launched a free tier targeting our mid-market segment
  Source: Sales team discussion in #product channel
  Relevance: competitive-positioning.md

## Action Items
- [ ] Update pricing page with new enterprise model
- [ ] Schedule competitive response meeting for Friday

## Problems
- Enterprise onboarding flow takes 3 weeks — customers complaining
  Severity: high
  Status: open — no resolution in conversation

Relationship to Other Skills