agentleFS
priyanshu-aryapriyanshu-arya/LinkedIn_Automation/.claude/skills/review-drafts/SKILL.md

review-drafts

Skill0 starsChanged 2 months agoMITView on GitHub

Use when the user wants to review, approve, or act on pending LinkedIn, X, or Substack drafts, or explicitly invokes /review-drafts. Presents each status:in_review note in Drafts/ (LinkedIn/X posts, X threads, Substack Notes, and Substack Articles alike) and collects Approve/Edit/Regenerate/Change Hook/Change Image/Change Time/Reject per REQUIREMENTS.md §12. This is the only skill allowed to move a draft to status:approved — nothing publishes or schedules from here.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If multiple drafts are pending, move to the next one after each decision.

    line 72
  2. Very specific:

    Only Approve and Reject end a draft's review.

    line 73
  3. Very specific:

    After each action, confirm what changed.

    line 79
  4. Very specific:

    Never auto-approve after Edit/Regenerate/Change Hook/Change Image/Change Time — those actions produce a revised draft that still needs an …

    line 93
  5. Distinctive:

    Never schedule, publish, or call Buffer from here.

    line 96
  6. Distinctive:

    Never silently drop a Reject reason — it's useful signal for future idea generation (a future phase could feed rejections back into …

    line 97

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Review Drafts (Approval Manager)

The human-in-the-loop gate for every platform. Nothing generated by earlier skills reaches Buffer or a manual Substack publish without passing through here. Genuinely platform-agnostic: it loops every status: in_review note in Drafts/ regardless of platform/type, and applies exactly one of the seven actions from REQUIREMENTS.md §12.

Arguments

  • draft-id (optional) — review one specific note. If omitted, go through every status: in_review note one at a time.

Process (per note)

1. Present

Show the platform (linkedin/x/substack-article/substack-note), then the content in whatever shape that note's type actually is:

  • draft (LinkedIn, X singles/threads, Substack Notes): the full post text (all tweets in order if thread is non-empty), hashtags, character count.
  • substack-article: title, subtitle, and the full ## Article Body.

Then, for either shape: viral_score and its factor breakdown (from ## Critic Notes), linked visual brief(s) if any, and the assigned target date (target_date for a LinkedIn idea, or the matching platform_schedule entry for any other platform). If a ## Post Audit Notes section is present (written by /audit-draft, Phase 19 — an optional, annotation-only step that may or may not have run on a given note), show it alongside: the algorithm-mechanics findings with their confidence tiers and the Humanizer score report/caveats it carries. It's informational only — it never changes what actions are available below, and its absence just means /audit-draft hasn't been run on this note yet. Then ask which action to take, using these exact seven options (present via AskUserQuestion when interactive):

  • Approve — status: in_review → approved. Append {action: approved, date, note} to history. No other field changes.
  • Edit — ask the user for replacement text (plain chat reply, not a multiple-choice question — edits can be arbitrary length). Replace the ## Post Text section (or ## Article Body for a substack-article) with what they provide verbatim. Append {action: edited, date, note: "user-edited"}. Status stays in_review so the edited version gets a chance to be re-approved explicitly, not auto-approved by editing it.
  • Regenerate — re-run this note's platform-matched writer for its idea, replacing the content entirely: /write-draft (linkedin), /write-draft-x (x), /write-draft-substack-article (substack-article), or /write-draft-substack-note (substack-note). Append {action: regenerated, date}. Status stays in_review.
  • Change Hook — rewrite only the opening hook line(s), keep the rest of the body. Append {action: hook_changed, date}. Status stays in_review.
  • Change Image — re-run /generate-visual (linkedin/x/substack-note) or /generate-visual-substack-article (substack-article) for this note to produce a new brief, replacing the old visual_ids entry. Append {action: image_changed, date}.
  • Change Time — ask for a preferred HH:MM, set the note's preferred_time. Only meaningful for linkedin/x (Buffer-scheduled) — for a substack-article/substack-note, say plainly that publish timing is manual and this action doesn't apply. Append {action: time_changed, date, note}.
  • Reject — status: in_review → rejected. Ask for a one-line reason, store it in the history note. A rejected draft is never revisited automatically.

2. Loop

If multiple drafts are pending, move to the next one after each decision. Only Approve and Reject end a draft's review; every other action keeps it in_review for another look (possibly in the same session, possibly later).

Report back

After each action, confirm what changed. At the end of a run, summarize: how many approved, rejected, still in_review — broken out by platform if this run touched more than one. For LinkedIn and X specifically, also check the result against the rolling schedule requirement (REQUIREMENTS.md §10), per platform: if fewer than 2 days' worth of approved drafts exist ahead of today for that platform, say so explicitly. (Sending a notification for this is the Scheduler's job — schedule-approved/schedule-approved-x — this skill just surfaces the number honestly.) Substack has no Buffer queue to check — for substack-note/Substack Articles, just report the approved count; /publish-substack is what those approvals feed into next.

Hard rules

  • This is the only skill allowed to set status: approved.
  • Never auto-approve after Edit/Regenerate/Change Hook/Change Image/Change Time — those actions produce a revised draft that still needs an explicit Approve.
  • Never schedule, publish, or call Buffer from here.
  • Never silently drop a Reject reason — it's useful signal for future idea generation (a future phase could feed rejections back into dedup/fatigue detection).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.