The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/daily_pipeline.py executes the external lark-cli utility to transmit reports. While the use of subprocess.run with an argument list follows best practices to prevent shell-level injection, it relies on an external binary that must be pre-installed and authenticated by the user.
[METADATA_POISONING]: The presence of the .security-scan-passed file constitutes deceptive metadata. This file makes authoritative claims about a successful security scan to influence the analyzer's judgment, which is a suspicious behavior typically used to mask intent or bypass scrutiny.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect injection surface through its data processing pipeline:
Ingestion points: Real-time market data is fetched from the Sina Finance API within scripts/daily_pipeline.py.
Boundary markers: The skill does not use delimiters or ignore-instructions to isolate external data when building the markdown report.
Capability inventory: The skill has the capability to write to the local filesystem (JSON snapshots) and execute shell commands (lark-cli via subprocess.run).
Sanitization: There is no escaping or validation of the data retrieved from the external API before it is interpolated into the markdown report and sent to the Feishu platform.

pharma-daily-report