saeidzebardast

Claim you'll hear	What's closer to true right now	What to actually do
"AI can hack anything autonomously"	Largely AI-run intrusions are now documented (Anthropic), but humans still chose targets and approved key steps, and the model sometimes hallucinated results	Assume attackers have a tireless assistant; patch faster and reduce exposed surface
"Prompt injection will be solved soon"	OWASP says it's unclear whether fool-proof prevention exists (LLM01)	Design so a successfully injected agent still can't do much damage
"Agents will go rogue"	The realistic failure is an agent faithfully following instructions it shouldn't have trusted, using permissions you gave it	Least privilege with allowlisted tools, plus approvals on irreversible actions
"AI phishing is a future problem"	Fluent, personalized lures are cheap to produce today	Phishing-resistan

saeidzebardast

Field	Why it matters
Session and agent identity	Ties actions to a specific agent run, not just "the service account"
Input sources	Lets you trace a bad action back to an injected document or web page
Proposed tool call (raw)	Shows what the model asked for before validation changed anything
Validation result	Tells you whether your guardrails fired, and which one
Approver	Records whether a human or an automatic policy allowed it
Timestamp and outcome	Basic forensics

saeidzebardast

Good commit practice	Its counterpart in the issue	What the agent does with it
Meaningful, imperative subject	Outcome-focused title	Reuses it for the PR title and commit subject
Mention the issue number	Clear, linkable issue	Adds "Closes #123" and links the work back
Atomic commits	One issue, one change	Produces a small diff you can review
Link external resources	Pointers and context section	Reads the right code and docs first
Include tests in commits	Testable acceptance criteria	Writes the tests that prove the change
Use hooks for standards	Verification commands	Runs checks before asking for review
Be consistent	Shared issue template	Gets the same structure every time

saeidzebardast

Section	What it answers	Why the agent needs it
Title	What changes, in one line	Becomes the PR title and often the commit subject
Context	Why this matters now	Lets the agent choose between valid options
Current behavior	What happens today	Gives a baseline to reproduce and test against
Expected behavior	What should happen instead	Defines the target, not the implementation
Acceptance criteria	How we know it is done	Turns "done" into checks the agent can run
Scope and non-goals	What must not change	Prevents creative extra work in unrelated code
Pointers	Where to look in the code	Saves exploration time and wrong-file edits
Constraints	Rules that always apply	Security, performance, API compatibility, style

saeidzebardast

	Software Engineer	Product Engineer
Owns	The code and its quality	The product outcome
Main question	How do we build this well?	Should we build this at all?
Input	Tickets and specs from a PM	Users, usage data, own judgment
Success	A reliable, maintainable system	Users adopt it and come back
Depth	Deep in one or two layers	Broad, with depth in applying it
Best fit	Complex platforms, infrastructure	Fast-moving SaaS and consumer products

saeidzebardast

saeidzebardast

saeidzebardast

/**
 * Sort object by key
 * 
 * @param unordered object
 * @returns {Object} ordered object by key
 */
function sortObject(unordered) {
  var ordered = {};
  Object.keys(unordered).sort().forEach(function (key) {
    ordered[key] = unordered[key];

saeidzebardast

# Unity generated
Temp/
Library/

# Visual Studio / MonoDevelop generated
ExportedObj/
obj/
*.svd
*.userprefs
/*.csproj

saeidzebardast

options("/*",
        (request, response) -> {

            String accessControlRequestHeaders = request
                    .headers("Access-Control-Request-Headers");
            if (accessControlRequestHeaders != null) {
                response.header("Access-Control-Allow-Headers",
                        accessControlRequestHeaders);
            }