slack-channel-review

name

slack-channel-review

description

Manually triggered skill that analyzes Slack channel discussions using MCP tools, classifies them for playbook relevance, and auto-invokes packmind-update-playbook with structured findings. Use when wanting to mine Slack conversations for recurring patterns, conventions, decisions, or action items that should be captured in the Packmind playbook. Supports non-interactive mode via `CI=true` or `--non-interactive` for scheduled GitHub Actions runs.

Slack Channel Review Mining

Analyze Slack channel discussions from a configurable time window, classify messages for playbook relevance, and feed actionable findings into the packmind-update-playbook workflow.

Workflow: Phase 0 → 1 → 2 → 3 → 4 → 5. Follow every phase in order.

Prerequisite: The Slack MCP server must be available. Load references/slack-mcp-tools.md for tool signatures and parameters.

Phase 0: Collect Parameters

Gather the required inputs before fetching any data.

Execution Mode Detection

Determine the execution mode before collecting any parameters:

1. Check CI env var: run echo $CI. If the value is true, set non_interactive = true.
2. Check skill arguments: if the skill was invoked with --non-interactive, set non_interactive = true.
3. Parse optional arguments:
   • --days N — override the default time period (default: 7 days)
   • --channels channel1,channel2 — comma-separated list of channel names to analyze (default: general)

If neither condition is met, non_interactive = false (interactive mode — original behavior).

Collect Parameters

1. Gather channels:

   • Interactive: Ask the user which channels to analyze (default: general).
   • Non-interactive: Use --channels value if provided, otherwise default to general silently.

2. Gather time_period:

   • Interactive: Ask the user how far back to look (default: 7 days, max: 30 days).
   • Non-interactive: Use --days value if provided, otherwise default to 7 days silently.

   Compute the cutoff date as today - time_period in YYYY-MM-DD format.

3. Resolve channel IDs: Use mcp__slack__list_channels to map channel names to IDs. If a channel name is not found, report it and skip it.

4. Confirm before proceeding:

   • Interactive: Display and BLOCK until the user confirms:

     "Analyzing Slack channels [channel list] for the last N days (since YYYY-MM-DD). Proceed?"

   • Non-interactive: Log the parameters and proceed automatically:

     "Non-interactive mode: analyzing Slack channels [channel list] for the last N days (since YYYY-MM-DD)."

Phase 1: Fetch Channel Messages

For each resolved channel, use mcp__slack__search_messages to find messages within the time window.

• Use query: in:#channel-name after:YYYY-MM-DD
• Paginate to collect all matching messages.
• If no messages are found across all channels, inform the user and stop.
• Display progress: "Found N messages across M channels to analyze."

Phase 2: Fetch Thread Context

For messages that have thread replies (indicated by reply_count > 0 or thread metadata), use mcp__slack__read_thread to fetch the full thread context.

Filtering

Apply these filters to discard noise:

Low-value filtering — discard messages that:

• Are shorter than 15 characters (excluding emoji reactions)
• Match low-value patterns: consists only of emoji reactions, "ok", "thanks", "+1", "done", "ack"
• Are bot-generated status messages (deploy notifications, CI results) unless they triggered a substantive discussion

Thread deduplication — for threads, keep the root message plus replies that add substantive content. Discard simple acknowledgments within threads.

Display progress: "Retained N substantive messages (after filtering)."

If no messages remain after filtering, inform the user and stop.

Phase 3: Classify Discussions

Categorize each remaining message/thread by playbook relevance:

High relevance (keep)

• convention — Establishes or references a naming, structure, or organizational pattern (e.g., "we should always name our hooks useXxx")
• decision — Records an architectural or technical decision (e.g., "we're going with Option B for the auth system")
• best-practice — Recommends or discusses a better approach for correctness, performance, or maintainability
• recurring-pattern — Same topic/question appears across multiple discussions or channels
• action-item — Commits to a specific change that affects team workflows or coding practices

Low relevance (discard)

• social — General chat, greetings, off-topic
• status-update — Simple progress updates without technical substance
• question-only — Asks a question without reaching a conclusion or decision
• troubleshooting — One-off debugging help that doesn't generalize

Retain only High-relevance discussions.

Recurring Theme Detection

Group retained discussions by semantic similarity. A recurring theme is 2+ discussions across different channels or days that address the same underlying pattern. For each theme:

• Assign a short descriptive label
• List all contributing messages/threads
• Note the occurrence count

Codebase Relevance Gate

After classification, apply a strict codebase relevance filter to all high-relevance findings.

Litmus test: "Would an AI coding agent need to know this when writing, reviewing, or shipping code in this repository?" — if no, discard the signal.

Signal	Verdict	Why
"Always name hooks useXxx"	KEEP	Coding convention → standard
"Going with Option B for auth"	KEEP	Architecture decision → affects code structure
"Run nx test before pushing"	KEEP	Dev workflow → command
"PRs need one approval before merge"	KEEP	Affects how agent ships code
"Standup at 10am"	DISCARD	Meeting scheduling
"Retros bi-weekly"	DISCARD	Team ritual
"Use Jira instead of Linear"	DISCARD	Project management tooling
"Vacation via BambooHR"	DISCARD	HR process

Edge cases: When a signal straddles both worlds (e.g., "use Linear labels X, Y, Z for PRs"), apply the litmus test strictly — does it change how an AI agent writes or ships code? If not, discard.

Display progress: "Retained N codebase-relevant findings (discarded M team-life signals)."

Phase 4: Build Findings Report

Choose the output path based on execution mode:

• Interactive: .claude/tmp/slack-review-findings.md
• Non-interactive: .claude/reports/slack-review-findings-YYYY-MM-DD.md (using today's date, timestamped for CI artifact upload)

Create the target directory if needed.

Report Structure

# Slack Channel Review Findings Report

**Channels**: #channel1, #channel2
**Period**: YYYY-MM-DD to YYYY-MM-DD
**Messages analyzed**: N (after filtering)
**Actionable findings**: N
**Codebase-relevant findings**: N (of M high-relevance signals)

---

## Recurring Themes

### Theme: <label>
**Occurrences**: N across N channels/days

| Channel | Date | Author | Link | Summary |
|---------|------|--------|------|---------|
| #channel | YYYY-MM-DD | @author | [link](https://team.slack.com/archives/CHANNEL_ID/pTIMESTAMP) | Summary of message/thread |
| #channel | YYYY-MM-DD | @author | [link](https://team.slack.com/archives/CHANNEL_ID/pTIMESTAMP) | Summary of message/thread |

**Suggested playbook action**: <Create standard | Update standard X | Create skill | Create command | ...>
**Rationale**: <Why this theme warrants a playbook change>

---

## Individual Findings

### Finding: <short description>
- **Channel**: #channel-name
- **Date**: YYYY-MM-DD
- : @author
 : https://team.slack.com/archives/CHANNEL

URL Requirements

Every message/thread reference in the report MUST include the full, navigable Slack deep link (e.g., https://team.slack.com/archives/C01234ABC/p1234567890123456). This is critical because packmind-update-playbook extracts source URLs from the findings report for its packmind-cli diff --submit -m source attribution. The expected attribution format is:

<topic>: <summary> (source: Slack #channel https://team.slack.com/archives/CHANNEL_ID/pTIMESTAMP)

Construct the deep link from the channel ID and message timestamp: https://<workspace>.slack.com/archives/<CHANNEL_ID>/p<TIMESTAMP_WITHOUT_DOT> (remove the dot from the ts field, e.g., 1234567890.123456 → p1234567890123456). Never omit the thread link — always substitute the real URL.

Phase 5: Present and Hand Off

Interactive mode

1. Display a summary to the user:

   • Number of recurring themes found
   • Number of individual findings
   • List of suggested playbook actions

2. Ask the user:

   "Found N recurring themes and M individual findings. Review the full report at .claude/tmp/slack-review-findings.md. Proceed to update the playbook with these findings?"

3. On confirm: Invoke the packmind-update-playbook skill with the full report content as the intent. This maps to Case B (explicit intent) of that skill's Phase 0. Frame the intent as:

   "Update the Packmind playbook based on the following Slack channel review findings report: "

4. On decline: Inform the user the report is available at .claude/tmp/slack-review-findings.md for manual review.

Non-interactive mode

1. Display a summary (same as interactive: themes count, findings count, suggested actions).

2. Log the report output path:

   "Report written to .claude/reports/slack-review-findings-YYYY-MM-DD.md."

3. Invoke packmind-update-playbook with the full report content as the intent, passing along non-interactive mode. Frame the intent as:

   "Update the Packmind playbook based on the following Slack channel review findings report: "

   The packmind-update-playbook skill will auto-approve all changes and delegate creation of new artifacts to the corresponding creation skills (packmind-create-standard, packmind-create-command, packmind-create-skill), all running in non-interactive mode.

Resources

references/

• slack-mcp-tools.md — Slack MCP tool signatures and parameters used by this skill