The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill's workflow includes an explicit instruction to scan sensitive configuration files for the purpose of identifying technical constraints.\n- Evidence: Step 4 of the '工作流程' (Workflow) section in SKILL.md specifies scanning 'env' and 'config' files. These files frequently contain hardcoded secrets, API keys, or private credentials. If the agent summarizes or extracts data from these files into the 'docs/codebase-insight.md' report, it could lead to accidental credential exposure.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a broad attack surface for indirect prompt injection as it is designed to read and process the entirety of a project's source code, which is untrusted input.\n- Ingestion points: The skill reads directory structures, package managers (package.json, go.mod, etc.), route registrations, ORM models, and configuration files throughout the codebase as described in Steps 1 through 4 of the '全量扫描' (Full Scan) workflow in SKILL.md.\n- Boundary markers: There are no instructions in SKILL.md directing the agent to use delimiters or ignore instructions that may be embedded in comments or code within the analyzed files.\n- Capability inventory: The skill is granted access to the 'Bash' and 'Write' tools, allowing it to execute local commands and create files based on the content it processes.\n- Sanitization: The instructions lack any requirement for the agent to sanitize, escape, or filter external content extracted from the codebase before interpolating it into the final report.

codebase-insight