Instagram Photo Find

Find high-quality Instagram photos for any destination or place. Searches for Instagram posts via web search, downloads candidate images, vision-scores them...

Security Scan

The skill's instructions and requirements mostly match its stated purpose (finding and scoring Instagram photos); the only notable mismatches are minor (it assumes curl and local /tmp file writes but doesn't declare required binaries or mention cleanup/permissions).

Purpose & Capability

Name/description (find Instagram photos, download, vision-score, return sources) aligns with the runtime instructions. The actions (web search, download via Instagram media URL, vision scoring) are coherent with the stated goal.

Instruction Scope

Instructions specify web searches, downloading images using curl, saving files to /tmp, and running a vision model prompt — all relevant to the stated task. Minor concerns: SKILL.md assumes availability of curl and permission to write into /tmp (these are not declared), and it instructs returning local file paths which implies local storage of downloaded images (no cleanup or privacy/rights guidance is provided). Also notes that the /media/?size=l endpoint 'requires no authentication' — that may be inaccurate or unreliable depending on Instagram's current behavior.

Install Mechanism

No install spec and no code files (instruction-only). This is lowest-risk from an install perspective — nothing is written to disk by an install step.

Credentials

Skill requests no environment variables, credentials, or config paths. That is proportionate: it only needs network access and a vision model, so no secrets are requested.

Persistence & Privilege

always is false and the skill does not request persistent system changes or modify other skills/configs. It requires local file writes during runtime but does not request elevated privileges.

Assessment

This skill appears to do what it says: search Instagram posts, download candidate images, and rank them. Before installing, consider the following: (1) The SKILL.md assumes curl is available and that the agent can write to /tmp — but the metadata does not declare curl as a required binary. Ensure your agent environment provides curl (or update the skill metadata) and that writing to /tmp is acceptable. (2) Downloaded images are stored locally and the skill returns local file paths — think about storage, cleanup, and copyright/privacy implications (images may be copyrighted or contain personal data). (3) The /media/?size=l endpoint may be unreliable or blocked by Instagram; the instructions’ claim that it 'requires no authentication' may not hold. (4) If you allow autonomous invocation, remember the skill will perform network downloads and write files when called; this is normal for its purpose but widen the blast radius if combined with other risky capabilities. If any of these are unacceptable, ask the author to (a) declare curl as a required binary, (b) add explicit cleanup and privacy/copyright guidance, and (c) avoid returning raw local file paths unless you want downloaded images retained.

Find the best Instagram photo for a given destination or place name.

Workflow

Step 1 — Search for Instagram post URLs

Search Brave for: site:instagram.com/p/ {destination} popular photo

• Request 10 results
• If the destination is a specific attraction/restaurant, use its name directly (e.g. site:instagram.com/p/ "Barton Springs" Austin)

Step 2 — Filter candidates by title

From search results, pick the top 5-6 candidates based on titles that suggest scenic/visual content:

Prefer titles with:

• Visual/emotional language ("beautiful", "colors", "stunning", "golden hour", "aerial", "skyline")
• Specific landmark names
• Travel/photography language ("travel", "explore", "photography")

Skip titles that suggest:

• Brand/corporate posts (Nike, concerts, conferences)
• Personal diary posts ("photo dump", "my trip")
• Celebrity/influencer selfies
• Non-photo content (events, announcements)

Step 3 — Download images

For each candidate, extract the image via:

https://www.instagram.com/p/{shortcode}/media/?size=l

• Use curl -s -L (follows the 302 redirect to CDN)
• Skip any that return non-200 or < 10KB (likely broken/removed)
• Save to /tmp/ with descriptive names

Step 4 — Vision-score each image

Run each downloaded image through the vision model with this prompt:

Describe briefly. Rate 1-10 as a hero destination photo for {destination} (iconic, scenic, represents the place well). Description + score only.

Step 5 — Return results

Return images ranked by score. For each result, provide:

• Score and brief description
• Instagram post URL (instagram.com/p/{shortcode}/)
• Local file path

Parameters

• destination (required): Place name (e.g. "Mexico City", "Taormina Italy", "Barton Springs Austin")
• count (optional, default 2): How many top results to return
• search_variant (optional): Additional search terms to append (e.g. "photography", "travel guide")

Tips

• For broad destinations (cities/countries), the generic search works well
• For specific attractions, quote the name: "Palacio de Bellas Artes"
• If first search yields low scores (< 6), retry with landmark-specific queries
• The /media/?size=l endpoint requires no authentication
• Carousel posts return only the first image
• Video posts may return a thumbnail