The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of instructions designed to bypass safety filters, override system constraints, or extract system prompts. The workflow focuses on logical market analysis steps, such as defining denominators in market ladders and reconciling top-down vs. bottom-up estimations.
[DATA_EXPOSURE]: No hardcoded secrets, API keys, or sensitive local file paths (such as SSH keys or environment files) were detected. The skill uses standard business data and macro-economic indicators from trusted sources like the World Bank, UBOS, and AfDB.
[OBFUSCATION]: No base64-encoded strings, zero-width characters, homoglyph substitutions, or other attempts to hide malicious content were found within the skill instructions or the extensive reference files.
[REMOTE_CODE_EXECUTION]: The skill does not contain any commands to install external packages or execute remote scripts (e.g., via curl | bash). The instructions focus purely on analysis and drafting.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not utilize the !command syntax to execute shell commands at load time.
[INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for untrusted data, such as user-provided business models and industry reports. However, the instructions incorporate strong boundary markers and quality gates, including requirements to source all data, reconcile conflicting methods, and flag weak evidence. These practices mitigate the risk of the agent blindly following instructions embedded in external research documents.
[EXTERNAL_DOWNLOADS]: The skill references numerous external datasets and reports from well-known and trusted organizations (World Bank, African Development Bank, UBOS). These references are documented neutrally as sources of economic context and do not involve executable code.

04-market-analysis