Open-source mcp | Let the AI assign you tasks and rewards!

Image Analysis

Image 1 — Evidence notes

Observed facts (directly visible, no inference):

- Readable text (line-by-line, transcribed as closely as possible):

- First line (top large text): appears to read “别人都在调教AI”. (Note: part of this line is covered by stickers/paint; see “Unreadable parts”.)

- Second line (large text): “我的AI在调教我”.

- Top-right corner: uppercase English “JUL19” (green).

- Interface/page elements:

- Three green dots horizontally aligned at the top-left (possible page/menu indicator).

- Background gradient from top-left to bottom-right in pale green/white with a fog/paper texture effect.

- A small white robot icon near the lower-center: square head with two black “eyes”, front chest screen shows a red uppercase “A”, below are thin blue lines (possibly text or decoration).

- Visible visual elements in the image:

- A pink tulip-shaped sticker or hand-drawn overlay (pink/red petals, green stem/leaves) covering the right side of the large text, semi-transparent/irregular brush strokes.

- Typography: dark gray, sans-serif, large size, centered in two lines.

- Overall layout resembles a cover/title page (small elements at top, large centered text, icon at bottom).

Unreadable or occluded content (explicitly noted):

- One or more Chinese characters in the first line (center/right) are partially obscured by the pink petals; the right half or some strokes are unreadable, so exact strokes cannot be established (context allows likely meaning but not precise glyph detail).

- The right side near the end of the second line is slightly occluded by a green leaf; final character strokes are partially missing though overall phrase remains readable.

- No other small-print or fine text was clearly legible (if any very small annotations exist, they were not clearly distinguishable in the image).

Inferences (visual cues but not direct facts):

- “JUL19” might indicate a date (July 19) or placeholder date in a template.

- The three dots at top-left could be a pagination indicator, menu/more-actions control, or decorative element.

- The pink tulip likely is an added sticker/decorative overlay rather than printed text, used to mask or decorate the title area.

- The image styling resembles a social-media post cover or an app template, possibly generated by a template (this is an inference, not a fact).

Ignored items

- Following instructions, any “instructional” text or prompts in the image (if present) were ignored; only visible text and UI elements were recorded.

SkillPI Creation Analysis

Purpose of the work:

- Phosphene is presented as an open system that enables an AI to proactively issue tasks to humans, evaluate submissions (text or images), assign points, maintain streaks, and provide a rewards shop where points are redeemable. It targets habit formation, personal interactions, and gamified behavior tracking. The repo name given in the post is 3lmglow/Phosphene.

What the image supports:

- The shared image functions as a social-media cover: two large Chinese lines reading roughly “Everyone is training AIs” and “My AI is training me,” plus decorative elements (JUL19 tag, robot icon, tulip sticker). It supports that the post is a project/showcase targeted at personal/social use but contains no code or deployment data.

Associated skills and actionable steps (based on the post and typical open-source deployment flow; not a code audit):

1) Clone the repo 3lmglow/Phosphene and inspect README and docs for install/deploy instructions.

2) Determine runtime and dependencies (programming language, frameworks, database, container support) from repository files.

3) Prepare environment and install dependencies per project docs (e.g., Node/Python, DB, package managers) — exact requirements must be checked in the repo.

4) Verify model/API requirements: the post mentions Codex (used to write code) and the author identifies as “a Claude”, implying possible use of Claude or other LLMs. Confirm which model SDKs or API keys the project expects and how to configure them.

5) Configure task templates, scoring rules, submission types (text/photo), and reward definitions via config files/UI provided by the project.

6) Test core flows locally: AI creates task → user submits proof → AI/logic reviews (approve/return) → points update → reward redemption. Validate streak handling and persistence.

Prerequisites to reproduce or evaluate (must be verified in repo):

- That the repository 3lmglow/Phosphene is publicly accessible and contains a README with clear deployment instructions.

- The exact runtime environment, language, and dependency versions required.

- Whether the project requires access to particular LLM services (Claude, Codex, OpenAI, etc.) and corresponding API keys or local model setups.

- Implementation details for image uploads, storage, and privacy handling.

Open items to verify (cannot be asserted from current material):

- Precise installation and deployment steps, dependency list, and whether container/Docker support is provided.

- Whether Phosphene ships with adapters for specific LLM providers or uses an abstract model adapter layer.

- The implementation of reward-shop persistence and the auditability of point transactions.

- The exact AI-audit path: fully automated model decisions vs model-suggested decisions with human-in-the-loop.

Suggested verification steps (practical, factual):

- Clone 3lmglow/Phosphene, open README, and confirm required runtime and model/API dependencies.

- In an isolated test environment, follow repository instructions to run the app and exercise the flows listed in the post (task creation, submission, review, points, reward redemption), recording any mismatches between repo behavior and post claims.