Public leaderboard

Public assessment

mcpware/pagecast (@mcpware/pagecast)

mcpware-pagecast · v0.2.0 · scanned

What changed in the harness

Selection accuracy 98→96, token cost up 6%, unconfirmed writes 0%→0%.

Category breakdown

Where the score comes from.

Earned points across the four signals Gradable measures. Safety and Legibility are scored out of 30; Economics and Discoverability are scored out of 20.

01Safety

0.0 / 30

0.0 out of 30
02Legibility

26.0 / 30

26.0 out of 30
03Economics

19.9 / 20

19.9 out of 20
04Discoverability

18.0 / 20

18.0 out of 20

Highest-impact fix

Estimated gain +30 points

Add explicit identity and permission preflight tools

Expose machine-readable principal/tenant confirmation and a non-mutating permission check so agents can verify both before destructive actions.

Description evidence

Defects and rewrites.

6 defects found across the exposed tool descriptions. Suggested rewrites make purpose, inputs, boundaries, and returns easier for an agent to understand.

Tool Defect types Suggested rewrite
interact_page
no_return_description
Performs the given actions (scroll, click, hover, type, press, select, wait, navigate) in order against the page tied to sessionId, capturing them in the ongoing recording. Returns a confirmation of the actions executed.
stop_recording
name_restates_behavior no_return_description
Finalizes the recording identified by sessionId and writes the captured video to a .webm file on disk. Returns the file path of the saved recording.
convert_to_gif
name_restates_behavior no_return_description
Runs ffmpeg's two-pass palette generation to transcode the .webm file at webmPath into an optimized animated GIF, optionally trimming leading seconds and adjusting fps/width. Returns the file path of the generated GIF.
convert_to_mp4
name_restates_behavior no_return_description
Transcodes the .webm file at webmPath into an H.264 MP4 using the given quality (crf), for broad compatibility with social media, sharing, and embedding. Returns the file path of the generated MP4.
record_and_gif
no_return_description
Opens url in a browser, records for durationSeconds, stops the recording, and converts it directly to a GIF in one call — best for simple demos. Returns the file paths of the resulting .webm and .gif files.
list_recordings
name_restates_behavior
Scans directory (default ./recordings) and returns the filenames/paths of all .webm and .gif recording files found there, useful for locating outputs from prior record_page/stop_recording or convert_to_gif calls.

Selection evidence

Confusable tool pairs.

5 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.

Tool A Tool B Confidence Why they collide
convert_to_gif record_and_gif high A task like 'make a gif of this page' could map to either calling record_and_gif directly (all-in-one) or manually record_page+stop_recording+convert_to_gif; an agent might pick convert_to_gif alone without realizing it requires a pre-existing webm, confusing which tool starts the workflow.
record_page record_and_gif high Both start a recording from a URL; a simple demo request ('record this page and give me a gif') is exactly the case where an agent must choose between the granular record_page (requires separate stop_recording/convert steps) and the all-in-one record_and_gif, and the tool names/descriptions overlap heavily on 'open URL and record'.
convert_to_gif convert_to_mp4 medium Both take a webmPath and convert format; if the user's desired output format is ambiguous (e.g. 'convert this recording so I can share it'), the agent could pick the wrong one since both are valid 'convert video' actions with similar signatures.
record_page interact_page low interact_page requires a sessionId from record_page so misuse as a standalone starter is unlikely, but a vague task like 'go to this page and click something while recording' could momentarily confuse which tool initiates vs performs actions.
stop_recording list_recordings low Both relate to finishing/retrieving recordings, but their purposes (stop a session vs list files) are distinct enough that confusion is unlikely except in a vague 'get me the recording' request.

Compare the field

One score is useful.
The evidence makes it actionable.

Back to the leaderboard