The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (gather-git-and-docs.sh and gather-telemetry.sh) to perform its primary functions. These scripts utilize common system utilities such as git, find, grep, sed, and curl to aggregate project information and query telemetry APIs.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes external data from markdown files that are included in the agent's context. 1. Ingestion points: The gather-git-and-docs.sh script searches for and reads .md files in user-specified directories and retrieves contents from project memory files (~/.claude/projects/...). 2. Boundary markers: The script outputs clear markdown headers to demarcate gathered data, which helps the agent distinguish between sections but does not prevent malicious instruction following from within the data. 3. Capability inventory: The skill executes local shell scripts which have filesystem access and network capabilities. 4. Sanitization: No sanitization or escaping is performed on the text content read from markdown files before it is printed to stdout.
[DATA_EXFILTRATION]: The gather-telemetry.sh script transmits project usage data to Tinybird, a well-known data platform. 1. Credential handling: Authentication tokens (TB_TOKEN or TB_READ_TOKEN) are retrieved from environment variables or sourced from a .env.prod file located at the repository root. 2. Network destination: Data is sent to the Tinybird API endpoint (api.tinybird.co) to track session overview, agent usage, and skill frequency.

standup-context