Loading Ashlr Brief
The briefing
AI-generated illustration
Ashlr merged more than 90 pull requests on October 19. The themes: no key leaks through error messages, and Phantom Secrets setup you can preview and undo.
October 19 was the busiest day of the month in Ashlr's public repositories, with more than 90 merged pull requests across Phantom Secrets, ashlr-plugin, Locus, Ashlr MD, Lexicon and Phantom. Two themes ran through most of them. Every item below comes from a pull request title and description.
The first theme: an error message should never carry a secret. Some AI providers echo the key you sent back inside their error text. On the same afternoon, Phantom, Ashlr MD, Phantom Secrets, Locus, Lexicon and ashlr-plugin each merged a fix that keeps keys, proxy tokens and URL credentials out of the errors they show. The Phantom Secrets proxy also gained tests proving that an echoed key, even a partial one, never reaches the program that made the request.
A related set in ashlr-plugin stops a model from choosing where your key goes. One tool used to accept an endpoint URL as an argument. Now the endpoint comes only from the user's own settings, redirects are refused, and plain http is allowed only to the local machine. Several scripts also stopped passing tokens on the command line, where other programs can see them, and read them from standard input instead. Phantom made the same change for its Pulse token.
The second theme: setup you can trust. Phantom Secrets added phantom setup --dry-run, which shows the changes it would make to your agent clients without writing anything. It now refuses to overwrite a client entry that differs from its own unless you pass --replace, backs up what it changes and records a manifest. Then phantom setup --undo uses that manifest to put things back. The undo is a dry run by default and refuses any file that changed since setup.
Look at the undo design. It acts on a file only if the file's hash still matches what setup wrote. Anything else is named and left alone. If your tool edits user config, that rule is cheap to copy and saves people from a destructive undo.
Key leaks through error logs are a common and quiet failure. Ask any AI tool vendor whether provider error bodies are scrubbed before they are logged or shown. Six Ashlr tools added that protection on one day.
If you use Phantom Secrets, try phantom setup --dry-run on a spare project, then phantom setup --undo to see what it would restore.
Ashlr tools connected to this story. Review their source and setup requirements.
Delegate supported API work through scoped placeholders and a local proxy, keeping real keys out of agent context.
Source material behind this briefing. Our practical implications are Ashlr’s interpretation.
This briefing was prepared with AI assistance from the linked sources. The analysis may include interpretation; consult the original sources for complete context. Cover artwork is illustrative.
SecurityThe briefing
AI-generated illustration
AWS made its Continuum agent generally available. The service finds, validates, and proposes fixes for code vulnerabilities in repositories while checking the AWS environment.
SecurityThe briefing
SecurityThe briefing
AI-generated illustration
No pull requests merged in Ashlr's public repositories on October 20. Here is the plain-English idea behind the previous day's biggest security change.
Token-efficient tools and workflows for Codex and Claude Code. Read, search, edit, and inspect repositories with compact outputs and visible context usage.
AI-generated illustration
Nothing merged in Ashlr's public repositories on October 21. Here is the idea behind a phrase that runs through recent pull requests: value-free reports.