The Agent Skills Directory

[PROMPT_INJECTION]: The research synthesis pipeline processes untrusted external data, creating a surface for indirect prompt injection attacks.
Ingestion points: The workflow in references/synthesize-research-report/synthesize-research-report.md ingests raw interview transcripts, notes, and summaries provided by the user.
Boundary markers: The instructions for the synthesis phases (e.g., phases/phase1-familiarization.md) do not implement robust boundary markers or instructions to ignore embedded commands within the processed data.
Capability inventory: The skill possesses capabilities for shell command execution via scripts/cookiy.sh, local file system writes to the analysis/ directory, and network access to the vendor API.
Sanitization: There is no evidence of sanitization or filtering of external content before it is processed by the sub-agent pipeline.
[COMMAND_EXECUTION]: The skill includes and utilizes a standalone shell script to manage research operations via a command-line interface.
Evidence: The script scripts/cookiy.sh is used to interact with the Cookiy platform, employing standard utilities like curl, jq, sed, grep, and bash to handle JSON-RPC requests.
[EXTERNAL_DOWNLOADS]: The skill interacts with external vendor-managed endpoints to perform research-related tasks.
Evidence: The CLI interacts with the API at https://s-api.cookiy.ai for authentication, study management, and participant recruitment. The SKILL.md file also mentions fetching content from https://cookiy.ai/intro-for-agent.txt.

user-research-cookiy