SuperAPI AI Directory

instagram-comments

General & other

View repository

Updated 10-15 21:38

About this skill

Instagram comment management (instagram-comments)

Overview

View, reply to, and hide comments on published Instagram media. This skill accesses the Instagram Graph API's comments endpoint (manage_comments permission gate) to instagram_comments_list / instagram_comments_reply / instagram_comments_hide call the tool.

Permission: All three tools require the Meta app to have the manage_comments permission granted. Without the permission, the API rejects the request.

Trigger keywords

comment, reply, hide, moderation, spam, clean up

Workflow

Step 1: List comments

List the comments on a specific media (media_id):

instagram_comments_list(media_id="<미디어 ID>")
# → {"data": [{"id", "text", "username", "timestamp", "hidden": false}, ...]}
  • media_id 는 instagram_publish_image etc., such as the publishing tools return the media_id (or the media ID retrieved with instagram_insights ).

Step 2: Draft a reply

The reply to a specific comment is first written as a draft only . At this stage nothing is published yet.

Write the reply body to reflect the stored style profile as well (if available). For aggressive/spam comments, hiding is recommended instead of replying.

Step 3: Korean 3-stage audit (required before publishing a reply)

Replies go out publicly under the brand account name and remain on the business account as-is. Before showing them to the user , pass them through the ⟨Korean 3-stage audit⟩. The order is fixed:

  moai-coworker:ai-slop-reviewer     1차 일반 슬롭 정리
→ moai-writer:korean-spell-check     2차 맞춤법 — 제안 수집 (미공개 정보가 섞였으면 건너뜀)
→ moai-writer:korean-humanize        3차 정밀 윤문 + 맞춤법 반영 + Phase 6 최종 검수
  • [HARD] korean-humanizeis last. The exact artifact judged by the Phase 6 final reviewis what gets published.
  • korean-spell-checksends the original text to an external service (nara-speller.co.kr). If the reply contains information not yet public(unannounced release dates, non-public results, customer personal data), skip this step — even if omitted, korean-humanizechecks spelling as well. If skipped, note that fact in the result.
  • If the audit is judged as hold_and_report, do not publish. Show the reason as-is and return to Step 2.

Step 4: Approval gate (common to replies and hiding)

[HARD] Do not publish replies or hide comments without approval. Both happen in places customers can see.

  • [HARD] Obtain approval via the path in the §Approval request contract. Do not ask "Shall I post it?" in prose. Do not skip the gate on the grounds that you cannot ask — the choice of path follows the §Approval request contract.
  • [HARD] Do not summarize; show all the arguments that will actually be passed, exactly as they are:
What to show Reply Hide
Publishing account (IG_USER_ID) ✓ ✓
Target post media_id ✓ ✓
Target comment comment_id ✓ ✓
Target comment Full original textand author ✓ ✓
Audited reply full textand byte count ✓ —
One line on what changed in audit stage 3 ✓ —
Reason for hiding (spam / abuse / other) — ✓

The approval options are structured as follows:

Options 뜻
Execute as is (recommended) Publish or hide the reply exactly as the shown arguments specify
Rewrite / change target Return to stage 2 → pass audit stage 3 again → re-approve
Cancel Exit without executing anything
  • [HARD] Even when processing multiple comments at once, show them individually. Batch approvals like "hide 5 spam items" are prohibited — if one of the 5 is a legitimate customer complaint, the user cannot know. Present them all at once as a list, but each item's comment text, author, and reason must all be visible.

Hiding is reversible (stage 5 below). The reason approval is still required is that hiding is a decision to silence someone. Being reversible and being permissible are different. In particular, if a legitimate complaint misclassified as spam is hidden, the customer sees their comment disappear and leaves the brand.

Stage 5: Execution

Execute only what was approved.

# 답글
instagram_comments_reply(comment_id="<댓글 ID>", text="<감사를 마치고 승인된 답글>")
# → {"id": "<새 답글 ID>"}

# 숨김 (삭제가 아님 — 작성자와 페이지 관리자만 안 보이거나, 정책에 따라 전체 숨김)
instagram_comments_hide(comment_id="<댓글 ID>")
# → {"hidden": true}

Hiding is reversible — under Meta policy, it can be made visible again. Replies are not reversible.

  • [HARD] Do not automatically retry on failure. If replying fails ambiguously (timeout / no response), stop without retrying. The absence of a success signal is not evidence that the reply was not posted, and retrying without confirmation posts the same reply twice.
  • [HARD] This server has no means to verify whether a reply was posted. instagram_comments_list(media_id) only fetches that media's comment list— it does not query replies attached to a specific comment (instagram_api.py comments_list). Moreover, the endpoint path itself is still subject to verification (@MX:TODO). If this tool is used to determine "no reply," you will post twice without ever seeing the successful reply.
  • Therefore, after an ambiguous failure, ask the user to open the comment directly in the Instagram app or web to checkand only after receiving the user's confirmation that it was not posted execute again. The skill does not judge on its own.
  • If a tool to query replies is actually exposed later, this clause will be changed then. Do not write it in advance assuming such a tool exists.

Approval request contract (runtime-neutral)

[HARD] This skill's gate is not tied to a specific tool name. AskUserQuestionis merely a means of the Claude runtime, and other runtimes, including Codex, do not have that tool. If the contract is written in terms of a tool name, then in a runtime that lacks that tool the gate becomes a permanent blocker, and every task requiring approval simply stops. That is not safety, it is a breakdown.

Approval is obtained in the following order. From the top, the first one that is actually possibleis used.

The definition of approval is not the means but the result: showing the approval document to the user as is and receiving an explicit response to it. Below are the paths that produce that result, and the first possible one from the top is used.

Rank Path Condition
1 The runtime's structured question tool (AskUserQuestion 등) when that tool is exposed in the current session
2 Present the approval document in ordinary conversation and receive the response on the next turn when in direct conversation with the user. Even without the tool, this path is always open
3 Return a structured blocker → the higher-level orchestrator asks when running as a subagent

[HARD] The runtime's tool execution permission prompt is not approval. That prompt only asks "may I call this tool," and does not display the arguments, estimates, and consent items that the gate is supposed to show. Only when the entire approval document and the options are actually displayed is it recognized as path 2.

[HARD] Because path 2 exists, the situation of "having no means to ask" effectively does not exist. Wherever conversation is possible, the approval document can always be presented in writing. fail-closed applies only to fully unattended execution where neither conversation nor returning a blocker is possible— only in that case does it stop without executing.

[HARD] When using path 3, the blocker must itself be an approval request. If the higher level does not know what to ask, it cannot ask back, and then it deadlocks. Include all of the following:

  • The action to be approved, in one line (what is irreversible / how much is going out)
  • All arguments required by the gate (values not summarized)
  • List of options — in a form the higher level can present to the user as is
  • How to resume — upon receiving which answer, what is to be executed next

[HARD] In unattended execution where all three paths are impossible, do not execute (fail-closed). Having no means to ask does not mean approval was obtained. In this case, record "could not proceed because there was no means of approval" and stop — do not proceed silently. Conversely, stopping because there is no tool even though conversation is possible is also wrongIt is. Use route 2.

This contract CLAUDE.local.md §This is the gate-side application of the generality principle (identical behavior across 2 OS types × 2 runtime types). A gate that runs on only one runtime is considered incomplete.


Precautions

Situation Response
manage_comments No permission Permission must be added via Meta App Review
setup_required Error IG_ACCESS_TOKEN / IG_USER_ID Environment variable configuration
Personal account Graph API not supported — Professional account required
Endpoint unverified The comments endpoint path is subject to run-phase verification (@MX:TODO) — it is recommended to recheck the path in the official documentation on first use
At audit stage 3, hold_and_report determination Do not publish a reply. Show the reason as-is and return to stage 2
The reply contains undisclosed information or customer personal data korean-spell-check Omit (external transmission) — note the fact of omission in the result
The user cancels at the approval gate Exit without executing anything. No partial execution

Output format

## 댓글 관리 결과

**미디어**: <media_id>
**댓글 수**: N

| 댓글 ID | 작성자 | 본문 | 상태 |
|---|---|---|---|
| ... | ... | ... | 표시/숨김 |

**조치**: (답글/숨김 내역)

Pre-publish setup (first time only)

The same Instagram credentials as Threads (IG_ACCESS_TOKEN / IG_USER_ID) and additionally, in the Meta app, the manage_comments permission must be granted. For the issuance procedure, see mcp-servers/moai-mcp-threads-poster/CONNECTORS.md 's Instagram section.

Related skills

Skill When to use
instagram-post Publish post (publishing must precede comment management)

When not to use this skill

  • Publish post: instagram-post skill
  • View insights: instagram_insights Direct tool invocation
  • Threads comments: The Threads API is outside the scope of this plugin