PraisonAI versions prior to 4.6.78 contain an allowlist bypass vulnerability in its shell command execution component. Attackers can craft find commands that use the built-in -exec, -execdir, or -delete actions to read files that are otherwise blocked, delete arbitrary files, or execute binaries that are not on the allowlist. This bypass undermines the application’s shell metacharacter filters and enables the execution of restricted commands, thereby compromising confidentiality, integrity, and availability of the data the application manages.

The affected product is PraisonAI from MervinPraison; any release before v4.6.78 is vulnerable. No further version granularity is available in the vendor data, so any installation running an earlier release should be treated as at risk.

The CVSS score of 8.7 indicates high severity. The EPSS score is reported as less than 1 %, suggesting a low but non‑zero exploitation likelihood, and the vulnerability is not listed in the CISA KEV catalog. Based on the description, it is inferred that the attack vector involves an attacker influencing the parameters passed to find, either by submitting crafted input to PraisonAI or by tampering with the execution environment. Once the bypass is achieved, the attacker can execute arbitrary shell commands, read or delete protected files, and potentially compromise the entire system.