The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The script scripts/agent.py ingests data directly from Active Directory environments (such as object names, SIDs, and security descriptors) and incorporates this untrusted data into its final JSON report. While this is a necessary function for the skill's purpose, it creates a potential surface for indirect injection if an attacker can control AD object attributes.
Ingestion points: LDAP search entries retrieved via conn.search in scripts/agent.py.
Boundary markers: None explicitly used to separate external data from internal logic within the script flow.
Capability inventory: The script has the capability to write the resulting report to a user-specified file path using the --output CLI argument.
Sanitization: The script mitigates risk by using standard json.dumps for serialization, which prevents many common injection vectors into the final output file.
[DATA_EXPOSURE]: The skill performs network operations to connect to an Active Directory Domain Controller. It requires a domain username and password, which are passed as command-line arguments (--username and --password). While standard for such tools, this practice carries the risk of credentials appearing in process listings.
[REMOTE_CODE_EXECUTION]: The skill references the ldap3 Python package as a prerequisite. This is a well-known, legitimate library for LDAP operations and is appropriate for the skill's stated purpose.

analyzing-active-directory-acl-abuse