| name | ps:slack-answer |
| description | Fetch a Slack thread, research the current directory codebase to find the best answer, draft a reply, then ask the user to confirm before posting. Use when the user runs /ps:slack-answer <thread-url> or asks to answer/reply to a Slack thread. |
| argument-hint | <slack-thread-url> |
| allowed-tools | ["Bash","Read","Write","Agent"] |
Slack Thread Answerer
Arguments: $ARGUMENTS
Parse the Slack thread URL.
Step 0 — Bootstrap scripts if missing
ls "$HOME/.local/share/personal-skills/scripts/slack/fetch_slack_thread.py" 2>/dev/null || \
curl -fsSL https://raw.githubusercontent.com/minhtranin/personal-skills/main/install.sh | bash
Step 1 — Check tokens
bash "$HOME/.local/share/personal-skills/scripts/slack/check_slack_tokens.sh"
- Exit 0: continue.
- Exit 1: show the printed instructions to the user exactly as-is and stop.
Step 2 — Fetch thread
python3 "$HOME/.local/share/personal-skills/scripts/slack/fetch_slack_thread.py" "<URL>"
Outputs JSON: channel_name, thread_ts, parent (author, text), replies[] (author, ts, text, reactions), participants[], reply_count.
- If exit 1 (auth error): tell the user tokens expired — re-extract manually using F12 DevTools, then run
save_slack_tokens.py --token xoxc-... --cookie xoxd-....
- If exit 2: show error and stop.
Step 3 — Summarize the thread context
From the fetched thread, extract:
- The core question or problem — what is the thread asking for? (1–3 sentences)
- Key details — any constraints, error messages, code snippets, context already given in the thread
- What is NOT yet answered — identify the open question that still needs a response
Output this context block immediately:
## #<channel> — <thread topic>
**Replies:** <count> · **Participants:** <names> · **Date:** <date>
### What the thread is asking
...
### Context & details given
...
### Open question needing an answer
...
Step 4 — Research the codebase
Use the Explore agent to deeply research the current working directory to find the most relevant answer to the open question. Provide the full thread context (core question + details) as the research goal.
Search for:
- Source files, components, functions, or modules directly related to the topic
- README, docs, config files, or knowledge files (e.g.
docs/, knowledge/, *.md, *.txt)
- Recent changes (git log) that may explain the current state
- Tests or examples that demonstrate correct usage
The agent should return:
- The most relevant file(s) and code section(s) that answer the question
- A direct answer based on what was found
- Any caveats, limitations, or "it depends" nuances from the code
If the current directory has no relevant code (e.g. it's a config-only or empty dir), note that and proceed with general knowledge only.
Step 5 — Draft the reply
Using the research findings from Step 4, draft a Slack reply message that:
- Directly answers the open question in 2–5 sentences
- References specific code if found (file name, function name, or key concept — not full paths)
- Includes a code block if a snippet would help (use triple backticks)
- Is conversational — written as a helpful team member would reply in Slack, not as formal docs
- Ends with a follow-up if there are caveats or the answer depends on context
Show the draft reply to the user:
═══════════════════════════════════════════════════
DRAFT REPLY for #<channel> — <thread topic>
═══════════════════════════════════════════════════
<drafted reply text>
═══════════════════════════════════════════════════
Sources used: <comma-separated file names or "general knowledge">
Then ask: "Post this reply to the thread? (y/n/e=edit)"
Step 6 — Handle user response
If n — do nothing
Tell the user the reply was not posted. Offer: "Copy the draft above to paste manually."
If e — edit
Ask: "What would you like to change?" Apply the edit, show the updated draft, and ask again: "Post this updated reply? (y/n)"
If y — post the reply
Post the reply to the Slack thread:
python3 "$HOME/.local/share/personal-skills/scripts/slack/post_slack_reply.py" \
--url "<THREAD_URL>" \
--message "<REPLY_TEXT>"
- Exit 0: show
✓ Reply posted to #<channel>.
- Exit 1 (auth error): tell the user tokens expired, re-extract with F12.
- Exit 2: show error message and stop.
