Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

zhuocha

v1.3.0

招投标重复项目核实助手。当需要分析同一 reid 下的多条 jy_id 是否为真正重复时激活。数据源:dify_ns_re_readsource(reid分组) + dwd_bid_it_all(明细字段,5200端口) + ods_bid_content(原始正文,5100端口)。典型触发语:「找茬大师」「分析...

0· 42·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The skill claims to verify duplicate tender records using three internal data sources and to write results back to an internal result table. The SKILL.md and the Python script both target the same internal HTTP APIs on 192.168.88.51 and the same DB tables, which is consistent with the stated purpose. However, the SKILL.md insists the detailed fields must come from the dwd_bid_it_all table on port 5200, while the shipped script fetches those fields from TABLE_RE (dify_ns_re_readsource via 6100) instead of calling the 5200 API — a mismatch that indicates the implementation does not follow the documented data-source requirement.
!
Instruction Scope
SKILL.md instructs the agent/program to query internal HTTP APIs and to write JSON INSERTs into result.dify_ns_re_result; that matches the script. Concerns: (1) SKILL.md explicitly recommends using Python urllib and avoiding bash/curl for writing results — the included script uses requests (Python) which is fine, but the docs and code disagree on some SQL practice (SKILL.md says 'do not use LEFT() in SQL; fetch full detail and slice in Python', yet the script's get_detail uses LEFT(detail, N) in its SQL). (2) The docs state the API does not support DELETE/UPDATE and recommend TRUNCATE TABLE result.dify_ns_re_result to clear bad rows — TRUNCATE is a destructive, high-privilege operation and recommending it as the remedy is dangerous and disproportionate; the skill gives no guidance about backups, authorization, or safeguards. (3) Some SQL builders in the script can produce syntactically invalid queries when the 'done' set is empty (e.g., WHERE reid NOT IN ()), indicating fragile behavior that could produce unexpected errors or broad queries.
Install Mechanism
This is an instruction-only skill with a small Python script and no install specification. Nothing is written to disk by an installer step and no external packages are being pulled by an automated installer. Risk from install mechanism is low.
Credentials
The skill requests no environment variables or credentials, which is coherent. It does, however, perform network calls to internal IPs (192.168.88.51 on ports 5100/6100 and references 5200). Network access to internal APIs is expected for this use-case, but it means the skill will read and write internal DB data without any declared authentication. Ensure the target APIs require appropriate auth and logging; otherwise the script could read/write sensitive internal data unexpectedly.
Persistence & Privilege
The skill does not request permanent inclusion (always: false) and contains no code that modifies other skills or system-wide agent config. It writes to a result table in the target environment, which is expected functionality for this task.
What to consider before installing
This skill appears to implement the duplicate-tender checking workflow described, but there are important mismatches and risky instructions you should address before installing or running it: - Implementation vs docs mismatch: SKILL.md requires reading detailed fields from dwd_bid_it_all on port 5200, but the included script reads those fields from the dify_ns_re_readsource table via the 6100 endpoint. Confirm which data source is correct and update docs or code. Also the doc admonition to avoid LEFT() in SQL contradicts the script, which uses LEFT(detail, N). - Dangerous DB operation: The documentation suggests using TRUNCATE TABLE result.dify_ns_re_result to recover from bad inserts. TRUNCATE removes all data and requires high DB privileges — do NOT run this without backups, explicit authorization, and an audit of who can perform it. Ask the author for safer remediation steps (e.g., DELETE with WHERE + logged backup, or use transactions and a staging table). - Network / auth: The script posts and queries internal HTTP endpoints (no auth in code). Confirm these endpoints require authentication and that writing to result.dify_ns_re_result is allowed for the account/service that will run the skill. If these endpoints are unauthenticated, running the skill could leak or corrupt internal data. - Testing & fixes: Run the script in a safe, isolated environment (or against a staging copy of the APIs and DB) and verify behavior. Check/fix SQL generation when 'done' set is empty to avoid malformed queries. Verify the list of jy_id written to rd is correctly formed and that the payload encoding/charset is handled safely (the docs mention Chinese punctuation problems). If you cannot validate the above with the skill author or test environment, treat this skill as untrusted and avoid running it against production/internal services.

Like a lobster shell, security has layers — review code before you run it.

biddingvk977q944ahv79jvs8105gzmcv5844pk0data-analysisvk977q944ahv79jvs8105gzmcv5844pk0dedupvk977q944ahv79jvs8105gzmcv5844pk0latestvk977q944ahv79jvs8105gzmcv5844pk0

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments