Skills data-engineer Audit History
📦

Audit History

data-engineer - 4 audits

Version comparison

Capability and finding changes across audited versions, newest first.

VersionDateResultReview itemsChange vs previous
v4 LatestJul 23, 2026, 11:28 PM 1 confirmed0No capability change
v3 Jul 15, 2026, 03:03 PM 2 confirmed0No capability change
v2 Jul 15, 2026, 03:03 PM 2 confirmed0No capability change
v1 Jul 15, 2026, 03:03 PM 2 confirmed0Baseline

Jul 23, 2026, 11:28 PM

All 42 backtick detections are Markdown fences or inline code, not shell execution. Eight reconnaissance matches are schema identifiers, and multilingual diagrams explain the entropy alert. Destructive delete and overwrite examples still lack confirmation, backup, and environment safeguards.

1
Files scanned
1,058
Lines analyzed
2
Review items
0
False positives ignored

Confirmed security concerns (1)

Medium
Destructive Data Operations Lack Guardrails
The command reference shows an ALTER TABLE DELETE, and the PySpark template uses overwrite mode. Neither example requires confirmation, backup verification, or a dry run.
Both destructive operations appear directly in reusable examples, with no adjacent safety gate. The placeholder targets reduce immediate impact but not reuse risk.
Audited by: codex

Jul 15, 2026, 03:03 PM

All 51 static findings are false positives caused by Markdown formatting, SQL identifiers, multilingual text, and documentation examples. No prompt injection, hidden executable payload, or obfuscation was found. Operational examples can still delete or overwrite data without explicit confirmation safeguards.

1
Files scanned
1,058
Lines analyzed
3
Review items
0
False positives ignored

Confirmed security concerns (2)

Medium
Destructive Operations Lack Confirmation Safeguards
The command reference includes a database delete, and the PySpark template uses overwrite mode. The skill does not require confirmation, backups, or dry runs before these actions.
Both destructive patterns are explicit examples, and surrounding guidance directs the agent to deploy and run pipelines. Risk depends on access to live systems.
Low
Inline Credential Pattern in JDBC Example
The JDBC example places username and password values directly in pipeline code. Although placeholders, this pattern can encourage users to embed real credentials.
The username and password options are visibly embedded in the example. They are placeholders, so no actual secret exposure is present.

Risk Factors

⚙️ External commands (1)
Audited by: codex

Jul 15, 2026, 03:03 PM

All 51 static findings are false positives caused by Markdown formatting, SQL identifiers, multilingual text, and documentation examples. No prompt injection, hidden executable payload, or obfuscation was found. Operational examples can still delete or overwrite data without explicit confirmation safeguards.

1
Files scanned
1,058
Lines analyzed
3
Review items
0
False positives ignored

Confirmed security concerns (2)

Medium
Destructive Operations Lack Confirmation Safeguards
The command reference includes a database delete, and the PySpark template uses overwrite mode. The skill does not require confirmation, backups, or dry runs before these actions.
Both destructive patterns are explicit examples, and surrounding guidance directs the agent to deploy and run pipelines. Risk depends on access to live systems.
Low
Inline Credential Pattern in JDBC Example
The JDBC example places username and password values directly in pipeline code. Although placeholders, this pattern can encourage users to embed real credentials.
The username and password options are visibly embedded in the example. They are placeholders, so no actual secret exposure is present.

Risk Factors

⚙️ External commands (1)
Audited by: codex

Jul 15, 2026, 03:03 PM

All 51 static findings are false positives caused by Markdown formatting, SQL identifiers, multilingual text, and documentation examples. No prompt injection, hidden executable payload, or obfuscation was found. Operational examples can still delete or overwrite data without explicit confirmation safeguards.

1
Files scanned
1,058
Lines analyzed
3
Review items
0
False positives ignored

Confirmed security concerns (2)

Medium
Destructive Operations Lack Confirmation Safeguards
The command reference includes a database delete, and the PySpark template uses overwrite mode. The skill does not require confirmation, backups, or dry runs before these actions.
Both destructive patterns are explicit examples, and surrounding guidance directs the agent to deploy and run pipelines. Risk depends on access to live systems.
Low
Inline Credential Pattern in JDBC Example
The JDBC example places username and password values directly in pipeline code. Although placeholders, this pattern can encourage users to embed real credentials.
The username and password options are visibly embedded in the example. They are placeholders, so no actual secret exposure is present.

Risk Factors

⚙️ External commands (1)
Audited by: codex