hermes-tweet
Native Hermes Agent X/Twitter plugin for tweet search, account reads, monitoring, and approval-gated actions through Xquik.
hermes-agent
twitter
x
social-media
automation
SKILL.mdhermes_tweet/skills/hermes-tweet
GitHubHermes Tweet
Overview
Discover catalog-listed Xquik routes and run authenticated reads without guessed endpoints.
Private or state-changing operations require the action gate and user approval.
Enable tweet_action only after the user approves its endpoint, payload, account, and effects.
When to use Hermes Tweet
Use this Skill for X/Twitter research, social listening, monitoring, support triage, creator research, giveaway audits, community audits, and controlled publishing.
Use tweet_explore first when the user asks for a capability, endpoint, route,
or Xquik API route. Use tweet_read only after a read-only endpoint is known.
Use tweet_action only after the user requests a write, private read, monitor,
webhook, extraction job, giveaway draw, or media operation that requires action
permissions.
Prerequisites
- Install and enable the plugin with
hermes plugins install Xquik-dev/hermes-tweet --enable. - Review Hermes security-scan warnings. Dangerous plugins are blocked.
- Configure
XQUIK_API_KEYon the Hermes runtime host for authenticated reads.tweet_exploreremains available without the key or network access. - Leave
HERMES_TWEET_ENABLE_ACTIONSunset or false unless the workflow needs an approved write-like or private operation. - For project-local plugins, set
HERMES_ENABLE_PROJECT_PLUGINS=trueonly in a trusted repository. - Restart a gateway after environment changes and start a new session. Active
CLI sessions can use
/reload.
Permissions and capabilities
- Use
tweet_explore,tweet_read, andtweet_actiononly through the enabled Hermes Tweet toolset. - Network access is limited to catalog-listed Xquik API routes reached by those tools. Do not create direct HTTP fallbacks.
- Shell access is not part of normal operation. Use Hermes CLI commands only for the install and registry checks listed in Testing.
- Local file access is not part of normal operation. Do not write reports, credentials, logs, screenshots, or cached API payloads unless the user asks for an explicit export workflow.
- Environment access is limited to configuration presence checks for
XQUIK_API_KEY,HERMES_TWEET_ENABLE_ACTIONS, andHERMES_ENABLE_PROJECT_PLUGINS. Never request or echo their values. - MCP access is not required.
Instructions
- Confirm the plugin is enabled with
hermes plugins listand confirm tool registration withhermes tools list. - Use
tweet_exploreto find the catalog endpoint and method. - Use
tweet_readfor public read-only endpoints after the API key is configured. - Before
tweet_action, state the exact endpoint, payload, account, reason, and expected side effects, then get explicit approval. - Verify the tool response. Report policy, authentication, validation, or account errors without retrying through alternate routes.
Choose a tool
- For endpoint discovery, call
tweet_explorewith a short query. - For catalog-listed
GETroutes, calltweet_read. - For private or write-like routes, require enabled actions and user approval.
- When
tweet_actionis disabled, explain the environment gate. - When
XQUIK_API_KEYis missing, ask the user to configure it. - Never request the key value in chat.
- When Hermes shows
not enabled, runhermes plugins enable hermes-tweet. - Project-local plugins require
HERMES_ENABLE_PROJECT_PLUGINS=truein trusted repositories. - For unattended work, prefer
tweet_readand keep actions disabled. - Remote Desktop profiles run tools on the remote Hermes host.
- Keep dashboard-managed secrets in the runtime environment.
Safety
- Never ask for or reveal API keys, signing keys, passwords, cookies, or TOTP secrets.
- Never pass credentials in tool arguments.
- Use only catalog-listed
/api/v1/...endpoints. - Copied endpoint URLs are accepted only when they resolve to catalog-listed paths.
- Do not use account connection, re-authentication, API key, billing, credit top-up, or support-ticket endpoints.
- For posting, deleting, following, DMs, profile changes, monitors, webhooks, extraction jobs, and draws, summarize the action before calling
tweet_action.
Known risks and mitigations
- Broad requests. Start with
tweet_exploreand prefertweet_read. Require an approved endpoint and payload beforetweet_action. - Exposed secrets. Ask only whether environment configuration exists. Never request key values or pass credentials as tool arguments.
- Guessed endpoints. Accept only catalog-listed
/api/v1/...paths. Never create a direct HTTP fallback. - Account changes. Keep
HERMES_TWEET_ENABLE_ACTIONS=falseby default. Summarize each account-changing call before approval.
Output
- Return endpoint choices, result summaries, action previews, and fixes.
- Use concise Markdown and JSON-like Hermes Tweet payloads.
tweet_exploredoes not call the API.tweet_readperforms authenticated reads.tweet_actioncan change state only after explicit approval.
Error handling
Use the narrowest recovery step that preserves the read-first and action-gated contract.
- Missing tool. Confirm the plugin is enabled. Run
hermes tools list. - Missing API key. Configure
XQUIK_API_KEYon the runtime host without pasting its value into chat, then run/reloadin an active CLI session or runhermes gateway restartand start a new gateway session. - Unknown endpoint. Call
tweet_exploreagain. Never guess paths or create a direct HTTP fallback. - Disabled action. Keep it blocked unless the user requested it and
HERMES_TWEET_ENABLE_ACTIONS=trueis intentionally configured. - Request failure. Return the sanitized error and corrective step. Do not retry through another route.
- Missing slash command. Verify registration in an active Hermes session. Prompt text does not prove registration.
- Secret in input. Stop and ask the user to rotate it.
Examples
Search tweets.
{"query":"tweet search","method":"GET"}
Then call tweet_read with this payload.
{"path":"/api/v1/x/tweets/search","query":{"q":"AI agents","limit":25}}
Inspect trends.
Run /xtrends in an active Hermes session. Use tweet_explore when the task
needs a catalog endpoint or structured response instead of the slash command.
Post a tweet.
{"query":"post tweet","include_actions":true}
Then call tweet_action with this payload.
{"path":"/api/v1/x/tweets","method":"POST","body":{"account":"@example","text":"Hello from Hermes Tweet"},"reason":"Post the user-approved tweet."}
Testing
Run these checks after you install or upgrade the plugin in Hermes Agent.
- Run
hermes plugins enable hermes-tweetunless the install used--enable. - Run
hermes plugins listand confirm the plugin isenabled. - Run
hermes tools listand confirm thehermes-tweettoolset is enabled. - Confirm
tweet_exploreis available withoutXQUIK_API_KEY. - Confirm
tweet_readappears only whenXQUIK_API_KEYis configured. - Confirm
tweet_actionstays hidden or disabled unlessHERMES_TWEET_ENABLE_ACTIONS=true.
Use these CLI checks.
hermes plugins enable hermes-tweet
hermes tools list
Release trust gate
Complete these steps before you present this skill as NVIDIA-verified. Do the same before you call it ready for broad enterprise deployment.
- Run SkillSpector against the complete skill directory and resolve critical or high findings.
- Complete
skill-card.mdwith owner, license, use case, deployment geography, risks, references, output shape, and release version. - Include Tier-3 eval data and
BENCHMARK.mdfor the reviewed release. - Sign the exact reviewed skill directory and publish
skill.oms.sig. - Verify the published directory with the expected certificate chain.
Do not claim NVIDIA verification when those release artifacts are absent.