The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on shell commands for data management and reporting.\n
The SKILL.md and reference/cost-tracking-guide.md files include instructions for the agent to use cat, jq, mkdir, echo, and date to interact with local files.\n
These commands are used to initialize configurations, log transaction details, and calculate spend metrics, which aligns with the skill's stated purpose.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from local log files, creating a potential but low-risk attack surface.\n
Ingestion points: Reads cost data from ~/.claude/daily-cost.json and ~/.claude/cost-log.jsonl (referenced in SKILL.md).\n
Boundary markers: The skill uses jq to parse structured JSON data, providing implicit validation of the data format.\n
Capability inventory: Limited to shell command execution for file manipulation and reporting; no network or administrative capabilities are present.\n
Sanitization: The use of jq for filtering and aggregation ensures that the agent processes data according to a predefined schema.

cost-metering