Currently, the redaction engine (src/redact.js) uses static regular expressions to scrub standard credential signatures (such as ghp_ tokens).
We want to allow users to add custom target identifiers for redaction. To prevent developers from storing actual secret values inside configuration files, we will support only:
- Environment variable names to search and redact.
- Safe key-name patterns (e.g. redacting values of any key containing "secret").
- Optional custom regular-expression patterns.
Redaction applies only to the recorded evidence representation:
- stdout
- stderr
- stored command arguments
- environment metadata
ReproSpec must never modify or redact the actual command arguments before executing the captured command, because that could change program behavior.
Acceptance Criteria:
- The command is executed with its original arguments, while only the persisted evidence copy is redacted.
- Any invalid custom regular expression must be safely rejected during configuration validation.
- Matching values must be redacted from stdout, stderr, and command argument lists.
- No actual secret values may be written to the config file.
- Documentation must explicitly warn that redaction is best-effort and does not guarantee that every credential is caught.
Currently, the redaction engine (
src/redact.js) uses static regular expressions to scrub standard credential signatures (such as ghp_ tokens).We want to allow users to add custom target identifiers for redaction. To prevent developers from storing actual secret values inside configuration files, we will support only:
Redaction applies only to the recorded evidence representation:
ReproSpec must never modify or redact the actual command arguments before executing the captured command, because that could change program behavior.
Acceptance Criteria: