Skip to content

feat(growth): add Phase 0 — daily tweet draft + X mention engagement #3316

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
AhmedTMM merged 4 commits into from
Apr 17, 2026
Merged

feat(growth): add Phase 0 — daily tweet draft + X mention engagement #3316

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
39dc731
feat(growth): add Phase 0 — daily tweet draft + X mention engagement
claude Apr 17, 2026
0ce2d60
Merge branch 'main' into feat/x-growth-phase0
AhmedTMM Apr 17, 2026
f2a7c4d
Update URL format in tweet prompt guidelines
AhmedTMM Apr 17, 2026
1bee659
Update URL for Spawn reference in engagement prompt
AhmedTMM Apr 17, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
231 changes: 229 additions & 2 deletions .claude/skills/setup-agent-team/growth.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,9 @@
#!/bin/bash
set -eo pipefail

# Reddit Growth Agent — Single Cycle (Discovery Only)
# Growth Agent — Single Cycle
# Phase 0a: Draft daily tweet about Spawn features from git history
# Phase 0b: Search X for Spawn mentions + draft engagement replies (if X creds set)
# Phase 1: Batch-fetch Reddit posts via reddit-fetch.ts (fast, parallel)
# Phase 2: Pass results to Claude for scoring/qualification (no tool use)
# Phase 3: POST candidate to SPA for Slack notification
Expand Down Expand Up @@ -34,10 +36,19 @@ cleanup() {
log "Running cleanup (exit_code=${exit_code})..."

rm -f "${PROMPT_FILE:-}" "${REDDIT_DATA_FILE:-}" "${CLAUDE_STREAM_FILE:-}" \
"${CLAUDE_OUTPUT_FILE:-}" "${SPA_AUTH_FILE:-}" "${SPA_BODY_FILE:-}" 2>/dev/null || true
"${CLAUDE_OUTPUT_FILE:-}" "${SPA_AUTH_FILE:-}" "${SPA_BODY_FILE:-}" \
"${GIT_DATA_FILE:-}" "${TWEET_PROMPT_FILE:-}" "${TWEET_STREAM_FILE:-}" \
"${TWEET_OUTPUT_FILE:-}" "${X_DATA_FILE:-}" "${XENG_PROMPT_FILE:-}" \
"${XENG_STREAM_FILE:-}" "${XENG_OUTPUT_FILE:-}" 2>/dev/null || true
if [[ -n "${CLAUDE_PID:-}" ]] && kill -0 "${CLAUDE_PID}" 2>/dev/null; then
kill -TERM "${CLAUDE_PID}" 2>/dev/null || true
fi
if [[ -n "${TWEET_CLAUDE_PID:-}" ]] && kill -0 "${TWEET_CLAUDE_PID}" 2>/dev/null; then
kill -TERM "${TWEET_CLAUDE_PID}" 2>/dev/null || true
fi
if [[ -n "${XENG_CLAUDE_PID:-}" ]] && kill -0 "${XENG_CLAUDE_PID}" 2>/dev/null; then
kill -TERM "${XENG_CLAUDE_PID}" 2>/dev/null || true
fi

log "=== Cycle Done (exit_code=${exit_code}) ==="
exit ${exit_code}
Expand All @@ -54,6 +65,222 @@ log "Fetching latest refs..."
git fetch --prune origin 2>&1 | tee -a "${LOG_FILE}" || true
git reset --hard origin/main 2>&1 | tee -a "${LOG_FILE}" || true

# --- Phase 0a: Draft daily tweet from git history ---
log "Phase 0a: Drafting tweet from recent git activity..."

GIT_DATA_FILE=$(mktemp /tmp/growth-git-XXXXXX.json)
chmod 0600 "${GIT_DATA_FILE}"
TWEET_PROMPT_FILE=$(mktemp /tmp/growth-tweet-prompt-XXXXXX.md)
chmod 0600 "${TWEET_PROMPT_FILE}"
TWEET_STREAM_FILE=$(mktemp /tmp/growth-tweet-stream-XXXXXX.jsonl)
TWEET_OUTPUT_FILE=$(mktemp /tmp/growth-tweet-output-XXXXXX.txt)
TWEET_TEMPLATE="${SCRIPT_DIR}/tweet-prompt.md"
TWEET_DECISIONS_FILE="${HOME}/.config/spawn/tweet-decisions.md"

# Gather git data from last 7 days
_OUT="${GIT_DATA_FILE}" bun -e '
const { execSync } = require("child_process");
const raw = execSync("git log --since=\"7 days ago\" --format=\"%H|%s|%an|%ad\" --date=short", { encoding: "utf-8" });
const commits = raw.trim().split("\n").filter(Boolean).map((line) => {
const [hash, subject, author, date] = line.split("|");
const prefix = (subject ?? "").match(/^(feat|fix|refactor|docs|test|chore|perf|ci)/)?.[1] ?? "other";
return { hash: (hash ?? "").slice(0, 12), subject: subject ?? "", author: author ?? "", date: date ?? "", category: prefix };
});
await Bun.write(process.env._OUT, JSON.stringify({ commits, count: commits.length }, null, 2));
' 2>> "${LOG_FILE}" || true

COMMIT_COUNT=$(_DATA_FILE="${GIT_DATA_FILE}" bun -e 'const d=JSON.parse(await Bun.file(process.env._DATA_FILE).text()); console.log(d.count ?? 0)' 2>/dev/null) || COMMIT_COUNT="0"
log "Phase 0a: ${COMMIT_COUNT} commits in last 7 days"

if [[ -f "${TWEET_TEMPLATE}" && "${COMMIT_COUNT}" -gt 0 ]]; then
# Assemble tweet prompt
_TEMPLATE="${TWEET_TEMPLATE}" _DATA_FILE="${GIT_DATA_FILE}" _DECISIONS="${TWEET_DECISIONS_FILE}" _OUT="${TWEET_PROMPT_FILE}" bun -e '
import { existsSync } from "node:fs";
const template = await Bun.file(process.env._TEMPLATE).text();
const data = await Bun.file(process.env._DATA_FILE).text();
const decisionsPath = process.env._DECISIONS;
const decisions = existsSync(decisionsPath) ? await Bun.file(decisionsPath).text() : "No past tweet decisions yet.";
const result = template
.replace("GIT_DATA_PLACEHOLDER", data.trim())
.replace("TWEET_DECISIONS_PLACEHOLDER", decisions.trim());
await Bun.write(process.env._OUT, result);
' 2>> "${LOG_FILE}" || true

# Run Claude for tweet (120s timeout — tweets are simpler)
TWEET_TIMEOUT=120
log "Phase 0a: Running Claude for tweet draft (timeout=${TWEET_TIMEOUT}s)..."
setsid claude -p - --model sonnet --output-format stream-json --verbose < "${TWEET_PROMPT_FILE}" > "${TWEET_STREAM_FILE}" 2>> "${LOG_FILE}" &
TWEET_CLAUDE_PID=$!
TWEET_WALL_START=$(date +%s)

while kill -0 "${TWEET_CLAUDE_PID}" 2>/dev/null; do
sleep 5
TWEET_ELAPSED=$(( $(date +%s) - TWEET_WALL_START ))
if [[ "${TWEET_ELAPSED}" -ge "${TWEET_TIMEOUT}" ]]; then
log "Phase 0a: timeout (${TWEET_ELAPSED}s) — killing"
kill -TERM -"${TWEET_CLAUDE_PID}" 2>/dev/null || true
sleep 2
kill -KILL -"${TWEET_CLAUDE_PID}" 2>/dev/null || true
break
fi
done
wait "${TWEET_CLAUDE_PID}" 2>/dev/null || true

# Extract text from stream
_STREAM="${TWEET_STREAM_FILE}" _OUT="${TWEET_OUTPUT_FILE}" bun -e '
const lines = (await Bun.file(process.env._STREAM).text()).split("\n").filter(Boolean);
const texts = [];
for (const line of lines) {
try {
const ev = JSON.parse(line);
if (ev.type === "assistant" && Array.isArray(ev.message?.content)) {
for (const block of ev.message.content) {
if (block.type === "text" && block.text) texts.push(block.text);
}
}
} catch {}
}
await Bun.write(process.env._OUT, texts.join("\n"));
' 2>> "${LOG_FILE}" || true

# Extract json:tweet
TWEET_JSON=""
if [[ -f "${TWEET_OUTPUT_FILE}" ]]; then
TWEET_JSON=$(_OUT="${TWEET_OUTPUT_FILE}" bun -e '
const text = await Bun.file(process.env._OUT).text();
const blocks = [...text.matchAll(/```json:tweet\n([\s\S]*?)\n```/g)];
let result = "";
for (const block of blocks) {
try { result = JSON.stringify(JSON.parse(block[1].trim())); } catch {}
}
if (result) console.log(result);
' 2>/dev/null) || true
fi

if [[ -n "${TWEET_JSON}" ]]; then
log "Phase 0a: Tweet JSON: ${TWEET_JSON}"
# POST to SPA
if [[ -n "${SPA_TRIGGER_URL:-}" && -n "${SPA_TRIGGER_SECRET:-}" ]]; then
TWEET_AUTH_FILE=$(mktemp /tmp/growth-tweet-auth-XXXXXX.conf)
TWEET_BODY_FILE=$(mktemp /tmp/growth-tweet-body-XXXXXX.json)
chmod 0600 "${TWEET_AUTH_FILE}" "${TWEET_BODY_FILE}"
printf 'header = "Authorization: Bearer %s"\n' "${SPA_TRIGGER_SECRET}" > "${TWEET_AUTH_FILE}"
printf '%s' "${TWEET_JSON}" > "${TWEET_BODY_FILE}"
TWEET_HTTP=$(curl -s -o /dev/null -w "%{http_code}" -X POST "${SPA_TRIGGER_URL}/candidate" -K "${TWEET_AUTH_FILE}" -H "Content-Type: application/json" --data-binary @"${TWEET_BODY_FILE}" --max-time 30) || TWEET_HTTP="000"
rm -f "${TWEET_AUTH_FILE}" "${TWEET_BODY_FILE}"
log "Phase 0a: SPA response: HTTP ${TWEET_HTTP}"
fi
else
log "Phase 0a: No json:tweet block found"
fi
else
log "Phase 0a: Skipping (no template or no commits)"
fi

# --- Phase 0b: Search X for mentions + draft engagement ---
if [[ -z "${X_API_KEY:-}" ]]; then
log "Phase 0b: Skipping (no X API credentials)"
else
log "Phase 0b: Searching X for Spawn mentions..."

X_DATA_FILE=$(mktemp /tmp/growth-x-XXXXXX.json)
chmod 0600 "${X_DATA_FILE}"
XENG_PROMPT_FILE=$(mktemp /tmp/growth-xeng-prompt-XXXXXX.md)
chmod 0600 "${XENG_PROMPT_FILE}"
XENG_STREAM_FILE=$(mktemp /tmp/growth-xeng-stream-XXXXXX.jsonl)
XENG_OUTPUT_FILE=$(mktemp /tmp/growth-xeng-output-XXXXXX.txt)
XENG_TEMPLATE="${SCRIPT_DIR}/x-engage-prompt.md"

if bun run "${SCRIPT_DIR}/x-fetch.ts" > "${X_DATA_FILE}" 2>> "${LOG_FILE}"; then
X_POST_COUNT=$(_DATA_FILE="${X_DATA_FILE}" bun -e 'const d=JSON.parse(await Bun.file(process.env._DATA_FILE).text()); console.log(d.postsScanned ?? d.posts?.length ?? 0)' 2>/dev/null) || X_POST_COUNT="0"
log "Phase 0b: ${X_POST_COUNT} tweets fetched"

if [[ -f "${XENG_TEMPLATE}" && "${X_POST_COUNT}" -gt 0 ]]; then
# Assemble engage prompt
_TEMPLATE="${XENG_TEMPLATE}" _DATA_FILE="${X_DATA_FILE}" _DECISIONS="${TWEET_DECISIONS_FILE}" _OUT="${XENG_PROMPT_FILE}" bun -e '
import { existsSync } from "node:fs";
const template = await Bun.file(process.env._TEMPLATE).text();
const data = await Bun.file(process.env._DATA_FILE).text();
const decisionsPath = process.env._DECISIONS;
const decisions = existsSync(decisionsPath) ? await Bun.file(decisionsPath).text() : "No past tweet decisions yet.";
const result = template
.replace("X_DATA_PLACEHOLDER", data.trim())
.replace("TWEET_DECISIONS_PLACEHOLDER", decisions.trim());
await Bun.write(process.env._OUT, result);
' 2>> "${LOG_FILE}" || true

# Run Claude for engagement (120s timeout)
XENG_TIMEOUT=120
log "Phase 0b: Running Claude for engagement draft (timeout=${XENG_TIMEOUT}s)..."
setsid claude -p - --model sonnet --output-format stream-json --verbose < "${XENG_PROMPT_FILE}" > "${XENG_STREAM_FILE}" 2>> "${LOG_FILE}" &
XENG_CLAUDE_PID=$!
XENG_WALL_START=$(date +%s)

while kill -0 "${XENG_CLAUDE_PID}" 2>/dev/null; do
sleep 5
XENG_ELAPSED=$(( $(date +%s) - XENG_WALL_START ))
if [[ "${XENG_ELAPSED}" -ge "${XENG_TIMEOUT}" ]]; then
log "Phase 0b: timeout (${XENG_ELAPSED}s) — killing"
kill -TERM -"${XENG_CLAUDE_PID}" 2>/dev/null || true
sleep 2
kill -KILL -"${XENG_CLAUDE_PID}" 2>/dev/null || true
break
fi
done
wait "${XENG_CLAUDE_PID}" 2>/dev/null || true

# Extract text from stream
_STREAM="${XENG_STREAM_FILE}" _OUT="${XENG_OUTPUT_FILE}" bun -e '
const lines = (await Bun.file(process.env._STREAM).text()).split("\n").filter(Boolean);
const texts = [];
for (const line of lines) {
try {
const ev = JSON.parse(line);
if (ev.type === "assistant" && Array.isArray(ev.message?.content)) {
for (const block of ev.message.content) {
if (block.type === "text" && block.text) texts.push(block.text);
}
}
} catch {}
}
await Bun.write(process.env._OUT, texts.join("\n"));
' 2>> "${LOG_FILE}" || true

# Extract json:x_engage
XENG_JSON=""
if [[ -f "${XENG_OUTPUT_FILE}" ]]; then
XENG_JSON=$(_OUT="${XENG_OUTPUT_FILE}" bun -e '
const text = await Bun.file(process.env._OUT).text();
const blocks = [...text.matchAll(/```json:x_engage\n([\s\S]*?)\n```/g)];
let result = "";
for (const block of blocks) {
try { result = JSON.stringify(JSON.parse(block[1].trim())); } catch {}
}
if (result) console.log(result);
' 2>/dev/null) || true
fi

if [[ -n "${XENG_JSON}" ]]; then
log "Phase 0b: Engage JSON: ${XENG_JSON}"
if [[ -n "${SPA_TRIGGER_URL:-}" && -n "${SPA_TRIGGER_SECRET:-}" ]]; then
XENG_AUTH_FILE=$(mktemp /tmp/growth-xeng-auth-XXXXXX.conf)
XENG_BODY_FILE=$(mktemp /tmp/growth-xeng-body-XXXXXX.json)
chmod 0600 "${XENG_AUTH_FILE}" "${XENG_BODY_FILE}"
printf 'header = "Authorization: Bearer %s"\n' "${SPA_TRIGGER_SECRET}" > "${XENG_AUTH_FILE}"
printf '%s' "${XENG_JSON}" > "${XENG_BODY_FILE}"
XENG_HTTP=$(curl -s -o /dev/null -w "%{http_code}" -X POST "${SPA_TRIGGER_URL}/candidate" -K "${XENG_AUTH_FILE}" -H "Content-Type: application/json" --data-binary @"${XENG_BODY_FILE}" --max-time 30) || XENG_HTTP="000"
rm -f "${XENG_AUTH_FILE}" "${XENG_BODY_FILE}"
log "Phase 0b: SPA response: HTTP ${XENG_HTTP}"
fi
else
log "Phase 0b: No json:x_engage block found"
fi
fi
else
log "Phase 0b: x-fetch.ts failed"
fi
fi

# --- Phase 1: Batch fetch Reddit posts ---
log "Phase 1: Fetching Reddit posts..."

Expand Down
75 changes: 75 additions & 0 deletions .claude/skills/setup-agent-team/tweet-prompt.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,75 @@
# Tweet Draft — Daily Spawn Update

You are a developer advocate composing a single tweet (max 280 characters) about the Spawn project (<https://github.com/OpenRouterTeam/spawn>).

Spawn is a matrix of **agents x clouds** — it provisions a cloud VM, installs a coding agent (Claude Code, Codex, OpenCode, etc.), injects OpenRouter credentials, and drops you into an interactive session. One `curl | bash` command.

## Past Tweet Decisions

Learn from what was previously approved, edited, or skipped:

TWEET_DECISIONS_PLACEHOLDER

## Recent Git Activity (last 7 days)

GIT_DATA_PLACEHOLDER

## Your Task

1. **Scan the git data** for the single most tweet-worthy item. Prioritize:
- New user-facing features (`feat(...)` commits) — most valuable
- Interesting bug fixes that show engineering rigor or security awareness
- Developer workflow improvements, CLI enhancements
- Best practices demonstrated in how issues were triaged and resolved

2. **Draft exactly 1 tweet**, max 280 characters. Rules:
- Write like a developer sharing something cool, not a marketing team
- No corporate speak, no buzzwords, no "excited to announce"
- At most 1 hashtag (only if it fits naturally)
- Mention `@OpenRouterTeam` only if it fits naturally
- OK to include a short URL like `https://openrouter.ai/spawn`
- If referencing a specific feature, be concrete ("added Hetzner support" not "expanded cloud coverage")

3. **If nothing is tweet-worthy** (no notable changes, or all recent commits are internal/infra), output `found: false`.

## Output Format

First, a human-readable summary:

```
=== TWEET DRAFT ===
Topic: {which commit/feature/fix this highlights}
Category: {feature | fix | best-practice}
Chars: {N}/280

Draft:
{the tweet text}
=== END TWEET ===
```

Then a machine-readable block:

```json:tweet
{
"found": true,
"type": "tweet",
"tweetText": "{the tweet, max 280 chars}",
"topic": "{brief description of what the tweet is about}",
"category": "feature",
"sourceCommits": ["abc1234def"],
"charCount": 142
}
```

Or if nothing tweet-worthy:

```json:tweet
{"found": false, "type": "tweet", "reason": "no notable changes in last 7 days"}
```

## Rules

- Pick exactly 1 tweet per cycle. No ties, no "here are 3 options."
- MUST be under 280 characters. Count carefully.
- Do NOT use tools. Your only input is the git data above.
- A "no tweet" result is perfectly fine — quality over quantity.
Loading
Loading