generated: '2026-08-13'
method: searched
source: https://www.npmjs.com/package/@bufferapp/cli
note: >-
  These are Buffer's OWN agent skill files, not generated by API Evangelist.
  The first-party Buffer CLI (@bufferapp/cli, npm, v1.2.0) ships a `skills/`
  directory of markdown context files and installs them into an agent with
  `buffer install claude` (writes ~/.claude/skills/buffer/SKILL.md) or
  `buffer install codex` (appends a managed block to ~/.codex/AGENTS.md).
  `buffer context` prints them; `buffer context --list` lists the topics.
  Each file below is saved VERBATIM from the published package tarball
  https://registry.npmjs.org/@bufferapp/cli/-/cli-1.2.0.tgz — nothing was
  rewritten, summarised or added.
api: graphql/buffer-schema.graphql
surface: Buffer CLI (generated from the public Buffer GraphQL schema)
package_version: 1.2.0
package_published: '2026-07-13'
docs: https://developers.buffer.com/guides/cli.html
skills:
  - file: buffer-getting-started.md
    name: Getting Started
    source_file: skills/getting-started.md
    description: >-
      Invariants across every CLI command — JSON output contract, exit codes,
      schema discovery (`buffer schema list` / `describe`), input precedence,
      dry-run-before-send, `--fields` trimming, glossary, metrics, timezones.
    operations: [account, channels, posts, ideas, aggregatedPostMetrics]
  - file: buffer-workflows.md
    name: Workflows
    source_file: skills/workflows.md
    description: >-
      Canonical sequences — resolve timezone, list channels, inspect a channel
      before scheduling, create a scheduled post safely, create a post at a
      local time, create a draft, paginate a Relay connection, list idea
      groups, capture-then-promote an idea, check posting limits, delete a post.
    operations: [account, channels, channel, posts, post, createPost, createIdea, deletePost, ideaGroups, dailyPostingLimits]
  - file: buffer-pitfalls.md
    name: Pitfalls
    source_file: skills/pitfalls.md
    description: >-
      Things the schema accepts that still produce wrong behavior — share-mode
      vs scheduling-type semantics, ISO-8601 offsets, never guessing IDs,
      per-service minimum payloads for all eleven networks, thread first-item
      rule, union response reading, cross-org constraints, plan limits.
    operations: [createPost, editPost, channels, dailyPostingLimits]
  - file: buffer-rate-limits.md
    name: Rate Limits
    source_file: skills/rate-limits.md
    description: >-
      The three stacked windows (15m / 24h / 30d), the RateLimit and
      RateLimit-Policy headers plus the legacy X-RateLimit-* fallback, where
      remaining quota is reported, the 429 backoff policy, batching strategy,
      concurrency guidance, and which invocations do not count.
    operations: []
  - file: buffer-idempotency.md
    name: Idempotency
    source_file: skills/idempotency.md
    description: >-
      Retry safety per operation. States plainly that there is NO
      idempotency-key mechanism in the Buffer API today, gives a per-command
      idempotent/not-idempotent table, a retry matrix keyed on exit code, and
      an inspect-before-retry loop for createPost.
    operations: [createPost, createIdea, deletePost, posts, channels, account, dailyPostingLimits]
  - file: buffer-post-metrics.md
    name: Post Metrics
    source_file: skills/post-metrics.md
    description: >-
      Per-post vs aggregate access patterns, the normalized PostMetricType
      vocabulary and which values are deprecated until 2026-12-01, the
      cross-network aggregation rule, metricsUpdatedAt freshness semantics
      (~24h ingest lag), and defensive reporting workflows.
    operations: [posts, post, aggregatedPostMetrics]
x-evidence:
  fetched: '2026-08-13'
  registry: https://registry.npmjs.org/@bufferapp/cli
  tarball: https://registry.npmjs.org/@bufferapp/cli/-/cli-1.2.0.tgz
  http_status: 200