메뉴
Read and write data in a linked Blink project's database and file storage. Use to manage app data, query users, update records, upload/download files. Requires BLINK_PROJECT_ID secret to be set first.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Full LinkedIn automation — publish posts, comment, react, browse feed, search profiles, and check messages. Uses Blink OAuth for writes and session cookies for reads. Use when asked to post, comment, like, or read anything on LinkedIn.
Manage Google Ads campaigns, ad groups, ads, budgets, and reporting via the Google Ads API. Use when asked to list ad accounts, run GAQL reports, create campaigns, manage budgets, or pull spend / impressions / conversions data.
Manage Meta Ads (Facebook + Instagram) — ad accounts, campaigns, ad sets, ads, custom audiences, and insights via the Graph API. Use when asked to launch a Facebook/Instagram ad, list Meta ad spend, build a custom audience, or pull Meta campaign reports.
Manage LinkedIn Marketing API — sponsored campaigns, creatives, audiences, and ad analytics. Use when asked to launch a LinkedIn ad, list LinkedIn ad spend, build a LinkedIn matched audience, or pull LinkedIn campaign reports.
Manage Reddit Ads — campaigns, ad groups, creative, audiences, and reports. Use when asked to launch a Reddit promoted post, list Reddit ad spend, build a custom audience on Reddit, or pull Reddit Ads insights.
Read the user's Granola meeting notes — list meetings in a time window, fetch full notes/summaries/attendees for specific meetings, pull verbatim transcripts, and answer natural-language questions across the user's meeting corpus. Use whenever the user asks about meeting notes, what was discussed, decisions, follow-ups, action items, or transcripts. Requires a linked Granola connection (provider key `composio_granola`).
| name | blink-app |
| description | Read and write data in a linked Blink project's database and file storage. Use to manage app data, query users, update records, upload/download files. Requires BLINK_PROJECT_ID secret to be set first. |
| metadata | {"blink":{"requires_env":["BLINK_API_KEY","BLINK_AGENT_ID","BLINK_PROJECT_ID"]}} |
Access a linked Blink project's isolated SQLite database and file storage using blink db and blink storage.
Store your project ID as an agent secret:
blink secrets set BLINK_PROJECT_ID proj_xxx
Get your project ID from blink.new → your project → Settings → API Keys.
blink db query $BLINK_PROJECT_ID "SELECT * FROM users LIMIT 10"
blink db query $BLINK_PROJECT_ID "SELECT count(*) as total FROM orders WHERE status = 'pending'"
blink db query $BLINK_PROJECT_ID "INSERT INTO tasks (id, title, done) VALUES ('t_123', 'My task', 0)"
blink db exec $BLINK_PROJECT_ID schema.sql
blink db list $BLINK_PROJECT_ID
blink db list $BLINK_PROJECT_ID users
blink storage upload $BLINK_PROJECT_ID ./report.pdf --path reports/2024/report.pdf
blink storage list $BLINK_PROJECT_ID
blink storage list $BLINK_PROJECT_ID images/
blink storage url $BLINK_PROJECT_ID images/logo.png
blink storage download $BLINK_PROJECT_ID reports/2024/report.pdf ./local-report.pdf
blink storage delete $BLINK_PROJECT_ID old-file.pdf
blink db query <project_id> <sql>
blink db exec <project_id> <file.sql>
blink db list <project_id> [table]
blink storage upload <project_id> <local_file> [--path <storage_path>]
blink storage list <project_id> [prefix]
blink storage url <project_id> <storage_path>
blink storage download <project_id> <storage_path> [output_file]
blink storage delete <project_id> <storage_path>
All commands use $BLINK_PROJECT_ID from the agent secrets vault.