Reconciles data sources using stable identifiers (Pay Number, driving licence, driver card, and driver qualification card numbers), producing exception repor...
Security Analysis
high confidenceThe skill is an instruction-only reconciliation helper whose requested inputs and included artifacts match its stated purpose and it does not ask for credentials, installs, or unexpected permissions.
The name/description, SKILL.md, and the two included reference files all describe the same reconciliation task (matching by Pay Number and secondary document numbers, producing exceptions). There are no extra environment variables, binaries, or unrelated requirements that would be inconsistent with the described purpose.
Runtime instructions are narrowly scoped to normal reconciliation steps (normalize, validate keys, join, produce exception reports, stop-and-ask gates). The SKILL.md does not instruct the agent to read unrelated system files, environment variables, or contact external endpoints. It explicitly recommends read-only behavior and stopping for ambiguous cases.
No install spec and no code files are present (instruction-only). Nothing will be written to disk or downloaded by the skill itself, minimizing install risk.
The skill declares no required environment variables, credentials, or config paths. The reconciliation task legitimately requires access to input datasets (CSV/XLSX) but nothing else is requested.
always is false and the skill does not declare persistence or attempt to modify other skills or system-wide settings. Model invocation is enabled by default (normal) but without additional privileges or credential access this is not a concern.
Guidance
This skill appears coherent and instruction-only, but it will need access to whatever CSV/XLSX files you want reconciled. Before installing or running: (1) confirm the agent/process will only be given the datasets you intend to share (these files contain sensitive PII like pay numbers and driving licence numbers); (2) run the skill on a small anonymized sample first to verify outputs and reason codes; (3) ensure your environment or agent connector will not exfiltrate data to external services; and (4) review and adapt the exception reason codes and stop-gates to match your operational tolerances. If you require formal data-handling guarantees (retention, logging, audit), verify those at the platform/connector level because the skill itself does not enforce them.
Latest Release
v1.0.0
Initial release: Provides data reconciliation with exception reporting and "no silent failure" checks. - Matches datasets using stable identifiers (e.g., Pay Number, driver card, licence numbers). - Flags and categorizes unmatched records, mismatches, duplicates, and invalid entries, each with explicit reason codes. - Always produces an exceptions report in a standardized CSV format. - Includes configurable normalization, matching criteria, and user-defined thresholds for quality gates. - Stops workflows or prompts for user input when data mapping or tolerances are ambiguous.
More by @alvisdunlop
Published by @alvisdunlop on ClawHub
