The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/estimates.py script utilizes subprocess.run to execute the git rev-parse --show-toplevel command. This is used to programmatically determine the repository root directory for log file storage.
[DATA_EXFILTRATION]: The skill reads and writes to a local log file, estimates.tsv, located in either the project's docs/agents/ folder or the user's local data directory (~/.local/share/agent-estimates/). The data stored includes task IDs, types, estimation ranges, actual durations, and optional notes. No network exfiltration or transmission of sensitive credentials was observed.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted input via the --note and --type arguments in estimates.py, which are subsequently stored in a TSV file and printed during report generation. While this constitutes a vulnerability surface, the data is handled using the structured csv module and the impact is limited to local display.
[DYNAMIC_EXECUTION]: The instructions in references/agent-runtime.md and SKILL.md direct the agent to perform arithmetic using a calculator or by generating and running a short script. This is an intended functional behavior to ensure mathematical precision in estimates and does not involve executing external or untrusted code.

estimate